AN ABSTRACT OF THE DISSERTATION OF. Betsy A. Bancroft for the degree of Doctor of Philosophy in Zoology presented on May 10, 2007.

Transcription

1 AN ABSTRACT OF THE DISSERTATION OF Betsy A. Bancroft for the degree of Doctor of Philosophy in Zoology presented on May 10, Title: Ultraviolet Radiation as an Environmental Stressor of Amphibians Abstract approved: Andrew R. Blaustein My thesis explored the effects of and potential mediating mechanisms for an important environmental stressor, ultraviolet-b (UVB) radiation. UVB radiation has negative effects on organisms in both terrestrial and aquatic systems. I used meta-analysis to quantify the effects of UVB radiation on a diversity of aquatic organisms (Chapter 2). UVB negatively affects aquatic organisms by reducing both survival and growth. In particular, UVB reduces growth of embryos more than any other life history stage. Some taxonomic groups may be more affected by UVB radiation than others. In our analysis, the growth of members of the kingdom Protozoa was suppressed by UVB radiation to a greater degree than any other kingdom. These analyses suggest that UVB is an important stressor in both freshwater and marine systems. Amphibians are a common component of freshwater systems and are experiencing world-wide population declines. These declines may be due to a number of causes including habitat loss, introduced species, global climate change, disease, toxic chemicals and UVB radiation. I used meta-analytic techniques to quantify the effects of UVB radiation on amphibians. By synthesizing the results of 41 articles on

2 the effects of UVB radiation on amphibians (Chapter 3), I found a nearly 2-fold reduction in survival of amphibians exposed to UVB radiation. Salamanders (caudates) appear to be more susceptible to damage from UVB than frogs or toads (anurans). Moreover, survival of larvae was much lower than survival of embryos or metamorphic individuals under UVB radiation. In addition, I used factorial metaanalytic techniques to explore the interaction between UVB radiation and other stressors in amphibian habitats. UVB radiation acted synergistically with other stressors to reduce survival of amphibians. Behavioral avoidance of UVB radiation may help mediate the negative effects of UVB radiation on amphibians. In aquatic systems, behavioral avoidance usually requires movement out of shallow water, where UVB levels can be high, into deeper waters with lower UVB transmittance. However, these two microhabitats have very different thermal profiles, creating a trade-off between exploiting warm waters with high UVB levels and avoiding UVB by seeking cooler, deeper regions of ponds. I explored the microhabitat use of larvae of four species through a series of laboratory experiments, field experiments, and observational field transects at three different amphibian habitats (Chapter 4). Larvae did not avoid UVB radiation in either the laboratory or field experiments. Larvae in thermal gradients selected relatively high temperatures regardless of the UVB exposure at these temperatures. In field transects, salamander larvae were most common in deeper, cooler waters where UVB levels were lower. In contrast, anuran larvae were frequently observed in the warmer and shallower regions of each habitat. These regions also had the highest UVB levels,

3 suggesting that anuran larvae are exposed to high levels of UVB due to thermoregulatory behavior. Behavioral avoidance of UVB radiation is not the only mechanism amphibians may use to prevent damage from UVB. Pigments such as melanin may allow larvae to exploit warm shallow waters by absorbing harmful UVB radiation before it causes cellular damage. I tested the efficacy of melanin as a photoprotective pigment in the larvae of two species, Rana cascadae and Pseudacris regilla (Chapter 5). I found no evidence of a photoprotective function for melanin in these larvae. In contrast, lighter colored tadpoles grew more under UVB radiation compared to darker colored tadpoles. Overall, exposure to UVB reduced survival of P. regilla larvae and reduced growth of R. cascadae larvae. Larvae of both of these species were frequently observed in very shallow water with intense solar radiation. This thesis emphasizes the importance of UVB radiation as an environmental stressor in aquatic habitats. Many aquatic organisms are negatively affected by UVB exposure. My thesis work quantitatively demonstrated that UVB radiation is one factor that reduces survival of amphibians and suggests that some species are exposed to high levels of UVB radiation in natural habitats. While UVB radiation is not the sole cause of amphibian population declines, my work suggests that UVB radiation is an important stressor for amphibians that should not be overlooked. In addition, UVB radiation is clearly an important stressor for many other aquatic organisms. Future work should consider the effects of UVB in aquatic systems, particularly the effects of UVB radiation on community structure and ecosystem function.

4 Copyright by Betsy A. Bancroft May 10, 2007 All Rights Reserved

5 Ultraviolet Radiation as an Environmental Stressor of Amphibians by Betsy A. Bancroft A DISSERTATION submitted to Oregon State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Presented May 10, 2007 Commencement June 2008

6 Doctor of Philosophy dissertation of Betsy A. Bancroft presented on May 10, APPROVED: Major Professor, representing Zoology Chair of the Department of Zoology Dean of the Graduate School I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my dissertation to any reader upon request. Betsy A. Bancroft, Author

7 ACKNOWLEDGEMENTS I would like to thank my advisor, Andy Blaustein. I am grateful for the interest you always expressed in my ideas, even when they strayed outside the bounds of your expertise. Your support of all aspects of my thesis work, from meta-analyses to physiology, allowed me the freedom to explore my question from many angles. You ve been an excellent advisor, mentor, and friend. You taught me by example how to be a good scientist without taking myself too seriously. My committee provided help and support throughout my thesis. Bruce Menge provided countless letters and statistical advice on multiple aspects of this thesis. Dan Roby helped shape my views on conservation biology and ecophysiology. Virginia Weis provided guidance in physiological methods and is one of the best role models imaginable. Thank you for helping interpret messy gels and strange SOD results. Susan Tornquist was a very helpful and kind grad rep. Several other faculty members and staff in Zoology have contributed significantly to my experience as a graduate student at OSU. Bob Mason wrote many letters, offered advice on post docs and life, and gave me teaching opportunities that I am grateful for. Joe Beatty made my life as a teaching assistant and graduate student much easier through his constant support and sense of humor. Doug Warrick provided support throughout my time as a Human Anatomy and Physiology instructor. Elizabeth Borer read drafts of portions of this thesis and provided much needed help with meta-analysis techniques. Eric Seabloom provided statistical advice on several occasions. Traci Durrell-Khalife, Tara Bevandich, Sarah Cain, and Torri Schrock

8 have been extremely helpful and frequently went above and beyond the call of duty to help me out. Cindy Kent provided a friendly smile and the occasional piece of candy or chocolate. Morgan Packard provided help in all aspects of physiology lab work and made working in the lab fun. This dissertation would not have been possible without the support and friendship of members of the Blaustein lab. Erin Scheessele, John Romansic, Barbara Han, Catherine Searle, Lindsay Michael, Anna Jolles, Tiffany Garcia, and Josh Lawler made the Blaustein lab a fantastic home for five years. Erin introduced me to our field sites, taught me how to identify many amphibian species in Oregon, and helped me transition into life in Corvallis. Erin also taught me the value of chocolate, especially M&Ms, after a long day in the field. John has been an excellent sounding board for experimental design and statistical analysis. Barbara has been a good friend since the day she arrived. Our office has been the source of much laughter and many good conversations. Catherine and Lindsay, I wish you guys had arrived sooner! I am grateful to have overlapped with you two as long as I did. You are both amazing people who made my last two years in the Blaustein lab more fun than it should have been. Tiffany provided support, guidance, and fun in all aspects of life. Anna and Josh, your input has helped shape the way I see ecology. I am indebted to many graduate and undergraduate students for help and support throughout my time at OSU. In particular, my cohort (Chris Stallings, Laura Petes and Barbara Han) has been a constant source of support and friendship. I can t imagine doing this without you guys. Laura, I can t possibly thank you for everything

9 I should. You ve been there offering advice, laughter and friendship every single step of the way. I am so glad we got to go through the many stages of graduate school together. Nads, Tads and the Dictators forever! I would also like to thank Rocky Parker, Maria Kavanaugh, Erin and Evan Scheessele, Elise Granek, Angela Brandt, Joe Tyburczy, Doug DeGross, Jerod Sapp, Elisha Wood-Charlson, Santiago Perez, Angela Perez, Catherine Searle, Lindsay Michael, Dave Paoletti, and Katie Johnson for friendship along the way. Devon Quick, Amy Harwell, and John Howieson were excellent resources for A&P and provided advice about life in and outside of the classroom. Nick Baker, former undergraduate assistant, current grad student and friend, is a co-author on most of this dissertation and I am grateful for all of the volunteer hours (years!) he spent helping me with my projects. This dissertation would not have been possible without his help. Karen Tonsfeldt is a superstar undergraduate assistant and friend. Becky Hill helped on several projects that didn t make it into this document, but her help is greatly appreciated. The women from my indoor soccer team, the Misfits, are an incredible group and I am glad I got to run around the turf with each of them. I am also grateful for the love and support of my family. Don Matheson, Emily Sutherland, and Laura Petes are friends who have become family. Thank you for helping me maintain perspective on life and science. Emily, you have an amazing talent for finding humor in every situation. You have always been the friend that I can call in any mood and end up feeling better afterwards. I would like to thank Ed Fasy, my dad, for asking about my thesis every time I saw him and helping to collect

10 tadpoles for preliminary data in Chapter 4. Marilyn Stewart, my mom, is a source of love, encouragement and support. She also was a field assistant for data that do not appear in this dissertation. My sisters and their families, the Tilbys, Skyllings, Labrums, and Sachs, have made me the person I am and help remind me of what is really important in life. My nephew Jarrick Tilby gets special recognition for being one of the funniest people I ve ever met. I love all of you guys. I am grateful to Bill and Quan Bancroft, my in-laws, and Carrie Bancroft, my sister-in-law, for making me part of their family and offering love and support throughout this project. Last, but certainly not least, I would like to thank my husband Rob Bancroft who served as a field and lab assistant for every chapter of this thesis. I am incredibly lucky to have such an amazing, intelligent, and supportive partner. Thank you for all the meals, encouragement, laundry, patience, conversations, laughter and perspective you provided during this endeavor. I am sure you know more about amphibians and UV radiation than you ever imagined or dreamed possible! Thank you for listening to my ideas, helping with my projects, and believing in me every day.

11 CONTRIBUTION OF AUTHORS Nick J. Baker assisted with many aspects of this dissertation. He performed literature searches, extracted data and helped analyze data for Chapters 2 & 3. Nick also helped collect data both in the lab and in the field for Chapter 4. Catherine Searle helped collect field data for Chapter 4. Tiffany Garcia helped in the formulation of the idea, assisted in thermal gradient design, collected laboratory data, and provided housing for parts of Chapter 4.

12 TABLE OF CONTENTS Page 1 General Introduction... 1 Thesis organization Effects of UVB radiation in marine and freshwater organisms: a synthesis through meta-analysis... 7 Introduction 9 Materials and Methods Results.. 22 Discussion 26 3 A meta-analysis of the effects of ultraviolet B radiation and other stressors on survival in amphibians Introduction 64 Materials and Methods Results.. 71 Discussion 72 4 Larval amphibians seek warm temperatures and do not avoid harmful UVB radiation. 96 Introduction 98 Materials and Methods Results Discussion 112

13 TABLE OF CONTENTS (Continued) Page 5 Effects of skin color and UVB radiation on survival and growth of amphibian larvae. 127 Introduction 129 Materials and Methods Results Discussion Conclusions. 142 Bibliography Appendices. 170

14 LIST OF FIGURES Figure Page 2.1 Normal quantile plots of effect size for survival and growth analyses The effect of UVB radiation on survival The effect of UVB radiation on growth Effect of UVB radiation on growth in each kingdom Weighted histogram of effect sizes for all comparisons in the survival analysis and the growth analysis Weighted histogram of effect sizes after removal of effect sizes greater than 1 standard deviation from zero Conceptual framework for factorial meta-analysis Mean effect sizes (log response ratio) for all studies included in the analysis of the effect of UVB on survival alone Mean effect size estimates for the factorial meta-analysis UVB profiles from all sites during mid-day Distribution of the larvae of three species at mid-day on 20 July, 2006 in Susan s Pond Distribution of Bufo boreas larvae in Todd Lake on 10 August, Toad tadpoles at Todd Lake on 10 August Distribution of larvae at mid-day on 10 August 2006 at the Potholes Survival of Pseudacris regilla and Rana cascadae larvae after exposure to UVB radiation Growth of R. cascadae larvae in each treatment Average brightness of R. cascadae larvae in each treatment 140

15 LIST OF FIGURES (Continued) Figure Page 5.4 Relationship between growth and brightness for all larvae in the two UVB treatments. 141

16 LIST OF TABLES Table Page 2.1 Summary information for each comparison included in the survival analysis Summary information for each comparison included in the growth analysis Heterogeneity statistics for each model in the survival analysis Results from variance partitioning using taxonomic group in the survival analysis Relationship between mean effect size estimates and dose-rate variables in the survival analysis Heterogeneity statistics for each model in the growth analysis Results from variance partitioning using taxonomic group in the growth analysis Relationship between mean effect size estimates and dose-rate variables in the growth analysis Amphibian species, experimental variables and biological variables used in multiple linear regression analysis of the effect of UVB alone on survival Biological and methodological explanatory variables used in regression analysis of the effects of UVB on survival in amphibians Additional stressors used in factorial meta-analysis Final regression model after backwards selection Final regression models after excluding Blaustein laboratory studies Results from Mann-Whitney U test, equivalency test and power analysis in thermal gradient laboratory trials Depth and temperature correlations at Susan s Pond on 19 July 2006 at each observation time 119

17 LIST OF TABLES (Continued) Table Page 4.3 Correlation between water depth and mean percent animals observed for each species at Susan s Pond Correlation between water depth and mean percent Bufo boreas observed at each sampling time in Todd Lake.. 121

18 LIST OF APPENDICES Appendix Page A Criteria for inclusion in meta-analyses presented in Chapter B The effect of UVB radiation on survival (log response ratio [lnr]). 172 C The effect of UVB radiation on growth (log response ratio [lnr]) D Amphibian species observed and transect information for each field site. 174 E Depth profiles for all ponds at the Potholes F Thermal contour plots for Susan s Pond on 20 July G Thermal profiles in Pond B and Pond C at the Potholes

19 Figure LIST OF APPENDIX FIGURES Page B The effect of UVB radiation on survival (log response ratio [lnr]). 172 C The effect of UVB radiation on growth (log response ratio [lnr]) E Depth profiles for all ponds at the Potholes F Thermal contour plots for Susan s Pond on 20 July G Thermal profiles in Pond B and Pond C at the Potholes

20 DEDICATION This dissertation is dedicated to the memory of my uncle, Bruce Stewart. I miss him every day.

21 Ultraviolet Radiation as an Environmental Stressor of Amphibians CHAPTER 1: GENERAL INTRODUCTION Anthropogenic changes to the environment have altered the habitat of many organisms. Habitats have been altered with the addition of contaminants (Fleeger et al., 2003), introduction of exotic organisms (Sakai et al., 2001), alteration of flood (Gergel et al., 2005) and fire regimes (Allen et al., 2002), and increases in temperature (IPCC, 2007). In addition, reduction in the stratospheric ozone layer has resulted in increases in ultraviolet-b (UVB) radiation reaching the earth s surface (Kerr & McElroy, 1993; Madronich, 1993; Madronich et al., 1998; Solomon, 1999). Stratospheric ozone depletion and the concomitant increase in ultraviolet-b (UVB) radiation pose an important threat to ecological systems. Ultraviolet-B radiation negatively impacts organisms in both terrestrial and aquatic systems (Caldwell et al., 1998; Häder et al., 1998). At the organismal level, UVB damages DNA, proteins and lipids. This can result in increased mortality and effects on growth, development, photosynthesis and immunity (Tevini, 1993; Caldwell et al., 1998). These effects have been reported in many organisms including viruses and bacteria (Karentz et al., 1994), phytoplankton and algae (Häder et al., 2003), amphibians and fish (Blaustein et al., 1998; Hessen, 2003), crops and forests (Caldwell et al., 1998), crustaceans (Hessen, 2003) and humans (van der Leun & de Gruijl, 1993). The effects of UVB radiation on organisms in aquatic habitats are of particular concern for a number of reasons (Häder, 1993). UVB negatively affects many

22 2 freshwater and marine organisms (de Mora et al., 2000; Helbling & Zagarese, 2003). UVB causes mortality in zooplankton, amphibians and fish (Siebeck et al., 1994; de Mora et al., 2000; Blaustein & Belden, 2003; Hessen, 2003). Sublethal effects in aquatic organisms, such as reduced growth, behavioral changes and increased susceptibility to disease have also been reported after exposure to UVB radiation (e.g., Williamson et al., 1997; Belden et al., 2000; Salo et al., 2000). These effects may scale up to the population or community level, causing changes in community structure and function (Bothwell et al., 1994; Mostajir et al., 1999; Marinone et al., 2006; but see Wahl et al., 2004). Habitat alteration has negatively affected many aquatic organisms, including amphibians. Amphibians are of particular conservation concern as amphibian populations are declining more rapidly than either birds or mammals (Stuart et al., 2004) with perhaps as many as 122 species becoming extinct since 1980 (Mendelson et al., 2006). Many factors appear to be contributing to amphibian population declines. These include habitat loss, pathogens, contaminants, climate change, and increases in UVB radiation (Blaustein & Kiesecker, 2002; Collins & Storfer, 2003). The effects of UVB radiation on amphibians include increased mortality, reduced growth, developmental abnormalities, increased susceptibility to disease and behavioural changes (reviewed in Blaustein et al., 1998; Blaustein & Kiesecker, 2002). In addition, the magnitude of the effect may vary among species or among different populations or life history stages of the same species (e.g., Belden & Blaustein, 2002a). For example, survival of moor frog (Rana arvalis) embryos was

23 3 higher when UVB radiation was filtered out compared with embryos exposed to UVB radiation (Häkkinen et al., 2001). However, survival of larval moor frogs was not affected by UVB exposure (Häkkinen et al., 2001). In contrast, survival of embryonic common toads (Bufo bufo) was not affected when exposed to UVB, but survival was lower in larvae exposed to UVB compared with those shielded from UVB (Häkkinen et al., 2001). Many studies suggest that UVB may interact synergistically with other stressors such as contaminants, climatic factors or pathogens (Blaustein et al., 2001, 2003). For example, western toad (Bufo boreas) embryos are susceptible to a complex interaction between UVB radiation, a pathogenic water mould (Saprolegnia sp.), and changes in precipitation (Kiesecker et al., 2001). Thus, mortality in western toad embryos increases when they are infected with Saprolegnia in the presence of increasing UVB radiation during years of lower precipitation when the UVB shielding property of the water is diminished. Climate cycles such as El Niño and La Niña, and the associated changes in precipitation, affect this dynamic (Kiesecker et al., 2001). UVB negatively affects amphibians both alone and in conjunction with other environmental factors. Amphibians have several defense mechanisms that may mediate the negative effects of UVB. Generally, amphibians either repair damage after it occurs or use various mechanisms to avoid or lessen exposure to UVB. Amphibians, like most organisms, use photorepair mechanisms (e.g., photolyase) to repair cyclobutane pyrimidine dimers in DNA induced by UVB exposure (Sancar & Sancar, 1988, Blaustein et al., 1994, Hays et al., 1996; Smith et al., 2002). However, there are

24 4 differences in the ability to repair UV-induced DNA damage. For example, in amphibians there are large differences in photolyase activity between species (Blaustein et al., 1994). Species with high levels of photolyase activity are usually more resistant to the harmful effects of UV radiation (Hays et al., 1996; Smith et al., 2002). Amphibians mitigate damage from UVB via two strategies: 1) behaviorally avoiding UVB radiation or 2) using photoprotective compounds such as the pigment melanin (Blaustein & Belden, 2003). Amphibians may avoid high UVB levels by seeking deeper waters (Belden et al., 2000; Blaustein & Belden, 2003; Licht, 2003). Thus, in choice experiments, larvae and adults of some species prefer areas with lower UV irradiance (Nagl & Hofer, 1997; van de Mortel & Buttemer, 1998; Belden et al., 2000; Garcia et al., 2004; Han et al., 2007). However, some species are frequently observed in warm shallow waters or basking in sunlight (e.g., Brattstrom & Warren, 1955; Lillywhite, 1970; O Hara, 1981; Bradford, 1984; Wollmuth et al., 1987), suggesting that these species do not avoid UVB radiation. Some species of larval amphibians darken in response to UVB (Belden & Blaustein, 2002b; Garcia et al., 2004). Darkening via photoprotective compounds such as melanin may allow certain species to exploit the resource-rich, warm shallows despite intense solar radiation and high levels of UVB present in these microhabitats. Mammals with darker skin are less prone to UV-induced skin damage than those with lighter skin (Kollias et al., 1991). However, darker skin does not appear to fully protect all amphibian species from lethal and sublethal UVB damage (Lesser et al., 2001; Belden & Blaustein, 2002b).

25 5 Thesis organization Many papers have been published on the effects of UVB on aquatic organisms, including several dozen on amphibians. There is a need for quantitative synthesis of the available information, both to resolve existing controversy and to generate further research directions. In Chapter 2, I used meta-analysis to quantitatively summarize the effect of UVB radiation on survival and growth in all aquatic organisms. This synthesis brings together two systems that have previously been considered separately in reviews and books: marine and freshwater systems. This broad analysis suggests that UVB reduces survival and growth in aquatic organisms. Although this analysis highlights the negative effect of UVB on aquatic organisms, the suggestion that UVB negatively affects amphibians has been controversial (e.g., Licht, 2003). In Chapter 3, I conducted a meta-analysis on the effects of UVB radiation on survival of amphibians. In addition, I used factorial meta-analysis techniques to explore the effects of UVB in conjunction with additional stressors on survival of amphibians. The two analyses in Chapter 3 reaffirm the large negative effect of UVB on survival of amphibians, and quantify the interaction between UVB and other common environmental stressors such as contaminants, pathogens, and ph. The strong negative effect of UVB detected in the meta-analyses suggests that UVB radiation is harmful to amphibians. Therefore, it is reasonable to predict that amphibians avoid exposure to UVB radiation in natural systems. Chapter 4 tests this

26 6 hypothesis using laboratory and field approaches. I tested microhabitat choice in a laboratory thermal gradient experiment and in a field UVB choice experiment, and then quantified habitat use in the field using transect surveys at three field sites. The results from these three approaches suggest that larval amphibians do not generally avoid solar radiation and exploit microhabitats where they are exposed to high levels of UVB. Amphibians exposed to high levels of UVB due to habitat use may be protected from UVB by photoprotective pigments such as epidermal melanin. If melanin acts as a photoprotective pigment, darker individuals should exhibit increased survival under UVB radiation compared to lighter individuals. I tested this hypothesis in Chapter 5 by manipulating larval amphibian skin color and exposure to UVB in the laboratory. The results from this chapter suggest that melanin is not an effective photoprotective pigment in all amphibian species.

27 7 Chapter 2 Effects of UVB radiation in marine and freshwater organisms: a synthesis through meta-analysis Betsy A. Bancroft, Nick J. Baker, and Andrew R. Blaustein Ecology Letters Volume 10(4):

28 8 Abstract Ultraviolet-B radiation (UVB) is a global stressor with potentially far-reaching ecological impacts. In the first quantitative analysis on the effects of UVB on aquatic organisms, we used meta-analytic techniques to explore the effects of UVB on survival and growth in freshwater and marine systems. Based on the large body of literature on the effects of UVB in aquatic systems, we predicted that UVB would have different effects in different habitats, experimental venues, trophic groups and life history stages. Contrary to our predictions, we found an overall negative effect of UVB on both survival and growth that crossed life histories, trophic groups, habitats and experimental venues. UVB had larger negative effects on growth in embryos compared with later life history stages. Despite the overall negative effect of UVB, effect sizes varied widely. In the survival analyses, no relationship between mean effect size and taxonomic groups or levels of exposure to UVB was detected. In the growth analyses, a larger negative effect on protozoans was observed. Our analyses suggest that the effects of UVB in aquatic systems are large and negative but highly variable between organisms. Variation in susceptibility may have important implications for population and community structure.

29 9 Introduction Stratospheric ozone depletion and the concomitant increase in ultraviolet-b (UVB) radiation pose an important threat to ecological systems. The work of Molina and Rowland (1974) on ozone-degrading compounds and the discovery of the Antarctic ozone hole (Farman et al., 1985) were influential in the formulation and signing of the Montreal Protocol on Substances that Deplete the Ozone Layer in This landmark international treaty was designed to halt the production and use of chlorofluorocarbons (CFCs) to avoid further degradation of the ozone layer. Despite the success of the Montreal Protocol (Blaustein et al., 2003; Solomon, 2004), the ozone layer remains damaged. This damage results in increasing levels of UVB radiation reaching the earth s surface (Kerr & McElroy, 1993; Madronich, 1993, 1994; Madronich et al., 1998; Solomon, 1999). Ultraviolet-B radiation negatively impacts organisms in both terrestrial and aquatic systems (Caldwell et al., 1998; Häder et al., 1998). Within living cells, nucleic acids, proteins and lipids are the primary targets of UVB damage (Buma et al., 2003). Cyclobutane pyrimidine dimers (CPDs) in DNA are the most common lesion induced by UVB exposure. These dimers as well as other types of UV-induced DNA damage can inhibit transcription and replication (Buma et al., 2003). Damage to proteins can interfere with normal cellular processes, while damage to lipids can disrupt cell membranes. Many organisms are able to repair damage caused by UVB radiation. Of these repair mechanisms, DNA repair is the most studied process. Organisms use different types of DNA repair mechanisms including photorepair,

30 10 excision repair and post-replication repair (Tevini, 1993). Photorepair is mediated by a group of enzymes called photolyases and relies on radiant energy in the UVA and photosynthetically active radiation (PAR) bands (Kelner, 1949; Sancar & Sancar, 1988). Photolyases primarily remove CPDs. Excision repair and post-replication repair are independent of radiant energy but are less efficient at repairing CPDs. Repair efficiencies can differ between species (e.g., Blaustein et al., 1994). Many experiments on the effects of UVB have been conducted in the laboratory (e.g., Damkaer et al., 1981; Charron et al., 2000; Roleda et al., 2004a). Although these experiments use ambient levels of UVB, researchers rarely replicate the full spectral environment found in natural systems. Because photorepair relies on wavelengths spanning UVA and PAR, laboratory experiments that do not carefully modulate the ratio of UVB to UVA and PAR may overemphasize the effects of UVB (Day & Neale, 2004). At the organismal level, damage to DNA, proteins and lipids may result in a variety of lethal and sublethal effects. These effects have been reported in many organisms including viruses and bacteria (Karentz et al., 1994), phytoplankton and algae (Häder et al., 2003), amphibians and fish (Blaustein et al., 1998; Hessen, 2003), crops and forests (Caldwell et al., 1998), crustaceans (Hessen, 2003) and humans (van der Leun & de Gruijl, 1993). The effects of UVB radiation on organisms in aquatic habitats are of particular concern (Häder, 1993). The effects of UVB on aquatic organisms depend in part on the dose of harmful radiation to which an individual organism is exposed. UVB dose is affected

31 11 by both organismal behaviour/location within the water column and the optical characteristics (i.e., UVB transmittance) of the water body. UVB penetration in aquatic habitats is modulated by factors including dissolved organic carbon (DOC), suspended particles (including phytoplankton), and surface reflection (Díaz et al., 2000; Hargreaves, 2003). In most aquatic habitats, UV attenuation is controlled primarily by absorption of UVB energy by DOC (Scully & Lean, 1994). As the majority of DOC is derived from terrestrial sources, freshwater habitats generally show higher UVB attenuation (and therefore lower levels of UVB in the water column) compared to marine waters (Kirk, 1994). Thus, marine organisms and freshwater organisms inhabiting clear alpine lakes may be most at risk to damage caused by recently increased UVB radiation. In some communities, organisms at lower levels in a food web tend to be more susceptible to environmental stress (Rafaelli, 2004). Aquatic autotrophs, including those inhabiting the shallow benthos, are generally exposed to some level of UVB. Several studies have shown negative effects of UVB on photosynthetic rates in aquatic systems (Litchman & Neale, 2005; Roleda et al., 2004a; Han et al., 2003). Unlike benthic autotrophs, phytoplankton are found throughout the water column and can be exposed to high levels of UVB (Villafañe et al., 2003). In addition, phytoplankton tend to be small in size and therefore have short pathlengths, allowing UVB penetration deep into the cells (Day & Neale, 2002). These observations have led to the speculation that phytoplankton may be particularly susceptible to damage from UVB radiation (Häder, 1993; Day & Neale, 2002; but see Halac et al., 1997). While

32 12 photosynthetic organisms must inhabit the photoactive zone, consumers (particularly mobile consumers) may behaviourally avoid regions of high UVB within the water column by using refugia or seeking deeper waters (reviewed in Leech & Johnsen, 2003). However, other constraints such as foraging behavior, avoiding predation, or thermal requirements could force consumers into areas with high UVB exposure (Leech & Johnsen, 2003). Consumers at these higher trophic levels in aquatic systems experience direct negative effects from UVB radiation (Siebeck et al., 1994; Williamson, 1995; Blaustein et al., 1998; Hessen, 2003). UVB causes mortality in some species of zooplankton, amphibians and fish (Siebeck et al., 1994; Blaustein & Belden, 2003; Hessen, 2003). Sublethal effects such as reduced growth, behavioral changes and increased susceptibility to disease have also been reported (e.g., Williamson et al., 1997; Belden et al., 2000; Salo et al., 2000). The effects of UVB can vary over life-history stage in both consumers and primary producers. Some life history stages may be less adept at repair or have habitat requirements that place the organism in areas with high UVB exposure (e.g., Siebeck et al., 1994; Epel et al., 1999; McNamara & Hill, 1999). For example, brine shrimp (Artemia franciscana) are more susceptible to damage from UVB in naupliar stages compared to the adult stage (Dattilio et al., 2005). Early life history stages of autotrophic organisms can also be more susceptible to damage from UVB. For example, early life history stages of some macroalgae are more susceptible to damage than the later stages (e.g., Roleda et al., 2004b). Embryos may be particularly

33 13 susceptible to damage from UVB due to reduced capacity for repair and limited mobility (Epel et al., 1999). Lethal and sublethal effects of UVB on both primary producers and consumers can cause shifts in community structure and function and thus can impact overall ecological processes in both aquatic and terrestrial systems (see reviews in de Mora et al., 2000; Hessen, 2003). For example, one half of all photosynthetic production is due to the activities of phytoplankton in aquatic systems (Houghton & Woodwell, 1989; Zepp, 2003). UVB radiation may alter community composition, size distribution, productivity, or nutritional quality of algae and photosynthetic bacteria (Karentz et al., 1994). Reductions in the biomass and photosynthetic yield of phytoplankton may reduce the available carbon sink in the oceans, resulting in higher atmospheric CO 2 concentrations (Zepp, 2003). Shifts in phytoplankton community composition, size distribution and nutritional quality could reduce the resources available to higher trophic levels. The potential of UVB to act as a stressor in aquatic environments has led to a number of reviews of the effects of UVB in both freshwater and marine systems (e.g., Häder, 1993; Karentz et al., 1994; Siebeck et al., 1994; Häder et al., 1998; de Mora et al., 2000; Helbling & Zagarese, 2003; Häder et al., 2003). However, organisms vary widely in their responses to UVB: some species are highly susceptible while others appear relatively resilient. Moreover, even within a species, susceptibility may differ between populations and in different life history stages. Many reviews focus on the

34 14 observed variation in susceptibility between organisms. A synthetic analysis of the overall effects of UVB is lacking. Previous reviews on the effects of UVB in aquatic systems have been qualitative. In general, these reviews present lists of effects and use the reported statistical significance of each study to assess the magnitude of the effects of UVB radiation. However, assessing the strength of an effect or importance of a stressor by counting the proportion of studies reporting a significant result (i.e., vote-counting ) has poor statistical power (Rosenberg et al., 2000). Meta-analysis techniques avoid the problems of conventional vote-counts and the subjectivity of traditional reviews. Meta-analyses have been used to identify broad trends in several aspects of global change (e.g., Root et al., 2003). We used meta-analytic techniques to test several hypotheses regarding the effects of UVB radiation in aquatic systems. For each hypothesis, we tested the effects of UVB on survival and growth separately. We hypothesized that UVB radiation has a negative effect on survival and growth of aquatic organisms. Furthermore, we hypothesized that the effect of UVB would be larger 1) in marine systems compared to freshwater habitats, 2) in the laboratory compared to field studies, 3) in primary producers compared to consumers, and 4) in earlier life history stages compared to later stages. Our paper is the first quantitative review of the effects of UVB in aquatic systems using rigorous statistical procedures. Our analyses reveal a large negative effect of UVB radiation on both survival and growth across all habitats, experimental venues, trophic groups, and life-history stages.

35 15 Materials and methods: Data selection We used six electronic databases (BIOSIS, Web of Science, Aquatic Sciences & Fisheries Abstracts, Fish & Fisheries Worldwide, Wildlife & Ecology Studies Worldwide, and Biological & Agricultural Index) to identify the studies used in our analyses. Within these databases, we searched for all combinations of the terms: ultraviolet, UV, UVB with survival, growth, and mortality to find primary literature on the effects of UVB radiation in aquatic organisms. We limited our search to experimental manipulations of UVB radiation (i.e., not UVA or UVA combined with UVB) where the investigators used the standard technique of applying plastic filters that differentially transmitted or filtered out UVB radiation. To explore both the lethal and the sublethal effects of UVB, we selected mortality and growth as response variables. Other sublethal response variables are possible (i.e., reproduction and lifestage duration) but growth is commonly measured for many organisms and can be assessed for multiple life stages. To avoid potential biases in the selection of studies, we established criteria for the inclusion of a study in the meta-analyses a priori (see Appendix A). Any data points within an article that met the criteria were considered for inclusion. To avoid personal bias, we made no attempt to judge study quality. Several articles included more than one species, location (e.g., lakes with different spectral transmission properties), UVB irradiance (or dose), or sampling period. All species and locations from a given article were included in our analyses if the overall criteria for including the study were met. Although including all species or

36 16 locations from one study might decrease the independence among some data points, the inclusion of all available species and environments allowed us to more fully explore the effects of UVB radiation in these systems (Gurevitch et al., 1992; Searles et al., 2001). However, if more than one ambient UVB irradiance or dose were used in the original article we randomly selected only one irradiance level or dose for inclusion. If the article reported survival or growth over a time series, we selected the final measurement for these analyses. When articles quantified growth using several response variables (i.e., length and mass), we randomly selected one variable for inclusion. All data were obtained from primary research articles. When necessary, data were extracted from published figures using TechDig V.2.0 software (Jones, 1998). We used the ITIS Catalogue of Life: 2005 Annual Checklist ( for taxonomic information. We followed the classification given by the authors for life-history stage; thus some categories may include photosynthetic organisms and non-photosynthetic organisms (e.g., embryo ) while other categories may be specific to certain taxa (e.g., sporophyte ). Effect sizes Our primary goal was to calculate an overall measure, including magnitude and direction (positive or negative), of the effect of UVB radiation on survival and growth in aquatic organisms. We used Hedges d as our metric of standardized effect size (Hedges & Olkin, 1985). Hedges d is an unbiased weighted measure of the difference between the means of the control and experimental groups divided by the

37 17 pooled standard deviation and multiplied by a correction term to adjust for small sample sizes. Convention dictates that a value of d greater than or equal to 0.8 is a large effect, d equal to 0.5 is a moderate effect, and d equal to 0.2 is a small effect (Cohen, 1969; Gurevitch et al., 1992). We defined the control group as the group shielded from UVB radiation; therefore, a negative value of d indicates a negative effect of UVB on survival or growth. We also calculated a log response ratio (lnr) as a measure of effect size, but because the results were qualitatively similar using both measures, we report only those based on d. The only differences between the two metrics are a larger effect of UVB on growth in primary producers compared to consumers, a larger (although not significantly larger) effect of UVB on growth in adult stages compared to embryos and larvae, and a nonsignificant effect of UVB on growth in field studies and on the larval life history stage (Appendix B and Appendix C). We used MetaWin Version 2.0 (Rosenberg et al., 2000) for all statistical procedures. We identified 115 articles with primary data on the effects of UVB radiation on survival. Of those, only 46 met our criteria, generating 87 total comparisons of 61 species (Table 2.1). Of the 87 comparisons, a significant difference in survival between UVB exposed and UVB shielded organisms was reported in 39 comparisons. We found 71 studies on the effects of UVB on growth in our searches. Only 27 of these met our criteria. The articles used in the analysis yielded 46 comparisons of 32 different species (Table 2.2). Out of the 46 comparisons, a significant difference in

38 18 growth between UVB-exposed organisms and UVB-shielded organisms was reported in 29 comparisons (Table 2.2). Full models We selected our response variables with the intention of quantifying both lethal and sublethal effects. As such, we used all survival data in one analysis and all growth data in a separate analysis. We used a random effects model to calculate the grand mean effect size for each analysis. Although fixed effects models are more typical in meta-analyses, we expected the true effect size to vary among studies due to the broad taxonomic scope of these analyses: thus, a random effects model was necessary (Gurevitch & Hedges, 1999). Random effects models use the pooled standard deviation to estimate the distribution of effect sizes within the population. Therefore, using a random effects model allows effect size estimates to vary not only due to sampling error, but due to real biological or environmental differences between organisms and studies. The output of each statistical test consisted of the grand mean effect size for the analysis with an accompanying bias-corrected bootstrapped 95% confidence interval (Adams et al., 1997) and a total heterogeneity statistic (Q). The mean effect size is significantly different from zero if the confidence intervals do not overlap with zero. The heterogeneity statistic is a weighted sum of squares and is tested against a chi-square distribution with n-1 degrees of freedom. A significant value of Q in a random effects model indicates that the variation among effect sizes is greater than expected from sampling error and the random component included in the model,

39 19 suggesting that all effect sizes may not come from the same population (Rosenberg et al., 2000). Exploratory analyses Our secondary goal was to examine the similarity in effect size among a priori selected groups including trophic group, habitat, life history stages, or experimental venue. We performed separate exploratory analyses testing the heterogeneity of mean effect sizes between groups using mixed effects models. This method of analysis is not ideal and is problematic in most cases. Performing multiple analyses on the same dataset increased the chance of Type I error. However, the purpose of these analyses was to quantify patterns in the literature, not to explain or partition heterogeneity. This distinction is subtle but important: future analyses should not use these methods to partition variance, but rather use hierarchical analytic methods (below, under Sources of heterogeneity). Due to low sample sizes in some groups, performing multiple analyses was the only way to explore the patterns in the literature. In addition, as groups with fewer than four comparisons were removed from the analyses, two of the four exploratory analyses used a subset of the data (trophic groups and developmental stages). We compared the mean effect size between marine and freshwater organisms, between studies conducted in the laboratory (and therefore under artificial UVB radiation) and studies conducted in the field (here defined as any experiment with natural solar radiation), between trophic groups, and between developmental stages. A mean effect size and bias-corrected bootstrapped 95% confidence interval were calculated for each group in the exploratory analyses. We

40 20 report parametric 95% confidence intervals when group sample size is small (i.e., <10) because parametric CIs provide a more conservative error estimate. When significant differences in mean effect size between three or more groups were observed in these exploratory analyses, we adjusted α using the Bonferroni method prior to conducting multiple two-way tests to determine where differences occurred (Gotelli & Ellison, 2004). Heterogeneity statistics were calculated to quantify both within-group (Q W ) and between-group (Q B ) variation. The interpretation of significant values of Q in mixed effects models is similar to random effects models. However, mixed effects models also incorporate group differences. In mixed effects models, significant values of Q suggest that observed variation is greater than expected due to sampling error, real variation in effect sizes (random component of model), and group differences (Gurevitch & Hedges, 1999). Although using only additive models may obscure some relationships (e.g., if embryos in freshwater habitats are more susceptible to damage from UVB compared to embryos in marine habitats our models would not detect the difference), current factorial meta-analytic techniques are only appropriate for use when the original studies are factorial in structure (Gurevitch et al., 2000). Sources of heterogeneity within groups We assumed that variation in effect size could be due to taxonomic grouping (i.e., closely related organisms may be more similar in effect size compared with more distant groups) or due to dose-rate for each experiment. We used a step-wise partitioning technique sensu Gurevitch et al. (1992) to identify sources of withingroup variation (Qw) using taxonomic information. Therefore, we used a mixed

41 21 effects model to compare mean effect sizes between taxonomic groups when enough comparisons ( > 4) were available. We used a continuous random effects model to explore the relationship between effect size and hours of UVB radiation per day, total hours of UVB for each experiment, and dose when available. Several weighting functions are commonly used to assess irradiance and dosage levels. Weighting functions are calculated according to which specific wavelengths in the UVB band are most damaging to particular organisms. These weighting functions include erythemal (McKinlay & Diffey, 1987), DNA (Setlow, 1974), and plant (Caldwell, 1971) action spectra. As dose estimates with different weighting functions are not comparable, each weighting function was considered separately. Sensitivity analysis For each analysis, including exploratory analyses, we used a form of sensitivity analysis to assess the influence of unusually large effect sizes on the analyses. We ranked each comparison by magnitude of effect size, removed each unusually large comparison step-wise, and re-ran each analysis. This procedure tests the influence of these large effect sizes on the conclusions of the analyses and the heterogeneity statistics. Publication bias For each analysis, we used several standard methods to identify potential publication bias ( file drawer problem; Rosenthal, 1979). We generated normal quantile plots of standardized effect size against the standard normal distribution to visually assess bias (Wang & Bushman, 1998; Rosenberg et al., 2000). We used

42 22 Spearman s rank correlation test to formally test for publication bias. In addition, we calculated Rosenberg s failsafe number (Rosenberg, 2005) to quantitatively assess the importance of potential publication bias to the outcome of our analyses. Rosenberg s failsafe number is the number of studies with an effect size of precisely zero necessary to change the results of an analysis from significant to non-significant. No evidence of publication bias was detected in normal quantile plots of standardized effect size (d) in either the survival analysis or the growth analysis, as the plots were relatively linear and all points fall within the confidence intervals (Figure 2.1). Spearman s rank correlation tests were nonsignificant for both survival (R = 0.087; p = 0.42) and growth (R = -0.04; p = 0.81), indicating no significant correlation between standardized effect size and sample size. Rosenberg s failsafe number was large for both the survival (227 comparisons) and growth (76 comparisons) analyses. Results Effect of UVB radiation on survival UVB radiation had a large negative effect on survival in aquatic organisms (Figure 2.2a). However, there were no differences in mean effect size between marine and freshwater organisms (Figure 2.2b), field and laboratory experiments (Figure 2.2c), trophic levels (Figure 2.2d), or life history stages (Figure 2.2e). With the exception of primary producers, UVB radiation had a large negative effect (d + > -0.8) in all groups (all effect size estimates were significantly different from zero). UVB

43 23 radiation had a moderate effect (d + = -0.6) on primary producers that was not different from zero. In all models, within-group heterogeneity was observed (Table 2.3). Therefore, we used hierarchical structure (taxonomic groups) and dose rate in an attempt to partition variance. Our analyses were limited by the small number of taxonomic groups with more than four comparisons; however, no differences between taxonomic groups were detected and significant within-group heterogeneity persisted through all levels of taxonomic grouping (Table 2.4). Similarly, significant residual error was found in each model examining dose-rate variables (Table 2.5). Surprisingly, the relationship between dose-rate variables and effect size was negative in only two models (hours of UVB radiation per day and DNA-weighted dose estimates). Effect of UVB radiation on growth UVB radiation had a large negative effect on growth in aquatic organisms (Figure 2.3a). No differences in growth were detected between habitats (Figure 2.3b) or experimental venue (Figure 2.3c). UVB radiation had a larger negative effect on primary producers than on consumers, but this trend was only marginally significant (Figure 2.3d; p = 0.057; Table 2.6). However, the effects of UVB radiation on growth varied by life history stage (Figure 2.3e; p = 0.002; Table 2.6). The effects of UVB radiation were larger in embryos than in larvae (p < ). No differences were detected between embryos and adults (p = 0.19) or between adults and larvae (p = 0.019; nonsignificant at α = after Bonferroni adjustment). The analysis

44 24 included only four studies of growth in embryos. The wide confidence intervals around the effect size estimate for embryos are a direct result of one large negative effect size (-27.55). A difference between these groups remained after removal of this large value (p = 0.02). As in the survival analysis, all mean effect sizes in the growth analysis were moderate to large and negative (d + > -0.7). Moreover, the mean effect size for every group was significantly different from zero. Although the between-group heterogeneity was frequently nonsignificant, the within-group heterogeneity was large in every model (Table 2.6). Similar to the survival analysis, we used taxonomic grouping and dose-rate variables to partition variance. UVB radiation had a larger effect on members of the kingdom Protozoa than any other kingdom in this analysis (Figure 2.4). However, heterogeneity within groups was still detected at the level of kingdom (Table 2.7). Further partitioning was limited by the number of comparisons within groups. Within-group heterogeneity persisted through all levels of taxonomic grouping (Table 2.7). The relationship between effect size and dose-rate variables was nonsignificant in all models except days of exposure (Table 2.8). Moreover, significant residual error persisted in all models except between dose and mean effect size (Table 2.8). Unfortunately, few studies reported dose and those that did used one of several possible weighting functions. Therefore, we are unable to fit different slopes using analysis of covariance models to assess the contribution of different taxonomic or other grouping variables on the dose-effect size relationship. Sensitivity analyses

45 25 Weighted histograms of effect size are left-skewed due to some extreme negative values (Figure 2.5). These values do not affect the overall normality of the data in the normal quantile plots (Figure 2.1). These two types of plots reveal different types of patterns in the data. In a weighted histogram, the height of each bar reflects the combined weight of the studies within that class (Rosenberg et al., 2000). If the effect sizes are plotted as a simple frequency histogram, the left skew is greatly reduced (data not shown). Therefore, the apparent conflict between the normal quantile plots and the weighted histograms is a direct result of the weighting function. The unusually large effect sizes have very low weights compared to the smaller effect sizes, even if their frequency of occurrence is similar. To test the robustness of our analyses against these extreme values, we removed each comparison with a large negative effect step-wise and re-ran each analysis. We began with the highest ranked effect size in each analysis and continued to remove the next largest effect size until Q was non-significant. Ten and eight comparisons were removed from the survival and growth analyses, respectively. After removal of these values, the weighted histograms were essentially unimodal with the highest frequency around zero effect size (Figure 2.6). Removing extreme values had no effect on the overall conclusions of the analyses: UVB radiation still had a large effect on survival and growth (d ++ = -0.88, -1.22), respectively. In addition, heterogeneity in the full model was reduced to non-significance after removal of these effect sizes (Q = 85.33, df = 73, p = 0.15; Q = 50.22, df = 37, p = 0.07) in the survival and growth analyses, respectively.

46 26 Removal of the largest values step-wise had few effects on the conclusions of the exploratory analyses. Removal of the 10 largest effect sizes had no effect on the conclusions of the survival analyses. Heterogeneity was reduced to non-significance in all survival analyses (data not shown). In the growth analyses, removal of the 8 largest effect sizes had no effect on the conclusions of the venue, habitat, or life history stage exploratory analyses. The difference in growth between trophic groups moved from marginally significant to non-significant when the largest effect size was removed from the analysis. Heterogeneity was reduced to non-significance in all but the analysis of the effects of growth between life history stages. Discussion UVB radiation had a large negative effect on both survival and growth in our analyses. Traditional reviews and syntheses of the effects of UVB on aquatic organisms generally suggest that the effects of UV vary widely among organisms (Siebeck et al., 1994; de Mora et al., 2000; Häder et al., 2003; Helbling & Zagarese, 2003). This assertion reflects the patterns in the literature. A conventional votecount of the comparisons included in the analysis would conclude that UVB radiation has a limited effect on survival. Less than half (45%) of the original comparisons observed a significant effect of UVB radiation on survival. A vote-count of the effects of UVB radiation on growth would detect a negative effect, as approximately 60% of studies reported a significant effect on growth. These analyses demonstrate the potential for meta-analytic techniques to identify broad trends that may be obscured by

47 27 variation and poor statistical power. Significance level depends on both the size of the measured effect and the sample size of each treatment (Gurevitch et al., 1992). Thus, two studies measuring the same effect may have different statistical outcomes simply due to different sample sizes. Due to the large heterogeneity statistics in both the survival and growth analyses, it may be incorrect to conclude that the grand mean effect size plus the 95% confidence interval is an accurate estimate of the true distribution of effects of UVB radiation in aquatic organisms (Hedges & Olkin, 1985; Gurevitch et al., 1992). However, when we removed the largest effect sizes, the heterogeneity statistic moved to non-significance. Thus, it is possible that there were two distinct populations of effect sizes that we did not isolate using taxonomic groupings or dose-response variables. To explore the heterogeneity further, we examined the details of each comparison with unusually large effect sizes to find commonalities between these comparisons that may indicate the source of the large effects. In the survival analyses, the large effect sizes were almost equally distributed across the groups (Table 2.1). All of the unusually large effect sizes were members of the kingdom Animalia (and therefore also classified as consumers in the trophic analysis), but this alone is not surprising given the paucity of comparisons in the other kingdoms (Table 2.1). The only striking similarity between these comparisons is that 9 of the 10 with the largest effect sizes also have extremely large variance (Table 2.1). In the growth analyses, 7 of the 8 unusually large effect sizes were laboratory studies. However, given that 40

48 28 of the 46 comparisons were laboratory studies, this was expected. All 8 of these comparisons were primary producers. These 8 comparisons represented members of all three phototrophic kingdoms (Table 2.2), but half of them were from the kingdom Protozoa. Similar to the survival analysis, all 8 comparisons removed from the growth analyses had extremely large variance (Table 2.2). Thus, the majority of the comparisons with unusually large effect sizes in each analysis also had large variance, and therefore may have low precision. Our estimates of variance were calculated using both sample size and the value of d for each comparison (Rosenberg et al., 2000). Samples with very large effect sizes and small sample sizes will therefore have large estimates of variance. However, small sample size alone did not always lead to a large effect size or a large variance. Considering the general lack of commonality between these comparisons we could not isolate the potential cause of the heterogeneity. Moreover, we could not justify removing these comparisons from the analysis. Therefore, below we discuss the results from the full analyses (with the very large effects included), but cautiously interpret results that were altered by the removal of the large effect sizes. We believe that for the majority of the comparisons in these analyses, the estimate of mean effect size and the associated 95% confidence interval is a reasonable approximation of the effect of UVB radiation on survival and growth in these organisms. Moreover, it is important to realize that the large heterogeneity observed in these models is driven by extremely large negative effects that may lie outside the expected distribution of effect sizes. These large effect sizes may be due to extreme experimental conditions or

49 29 alternatively, may indicate that these organisms are particularly susceptible to damage from UVB radiation. More research on these organisms with larger sample sizes is necessary to explore the basis for the large effect sizes observed in these comparisons. However, the large negative effect of UVB radiation on survival and growth persists with or without these unusually large effect size estimates. Our analyses suggest that the magnitude of the effect varies among organisms, but the effect tends to be large and negative. The negative effect of UVB on both survival and growth was detected despite different habitat types, life history stages, trophic levels and diverse taxonomic groups. No differences in survival or growth between marine and freshwater systems were detected, suggesting that the magnitude of the effect of UVB radiation cannot be predicted by habitat type. The optical qualities of marine waters can be very different from freshwater habitats. The depth at which 10% of surface UVB can be detected can vary by more than two orders of magnitude between temperate lakes and clear ocean waters, with much higher attenuation in freshwater habitats (Díaz et al., 2000). Despite the potential difference in UVB penetration between these systems, we did not observe a difference in effect size between freshwater and marine habitats. It is possible that the UVB penetration in these two habitat types was similar since the majority of marine studies were conducted in coastal environments where terrestrial run-off and estuarine contributions increase the levels of dissolved organic carbon in near-shore waters (Kirk, 1994).

50 30 Exposure of organisms to UVB in the laboratory rarely approximates natural environments. Of specific concern is the ratio of UVB to both UVA and PAR necessary for photorepair of DNA. In most laboratory experiments the UVB dose is closely monitored and applied, but the dose of UVA and PAR tends to be much lower than in natural systems (e.g., Ankley et al., 2000). Surprisingly, no difference in mean effect size between laboratory and field exposures was found in either the survival or growth analyses. We do not believe that the ratio of UVB to UVA and PAR is unimportant; rather, our analyses suggest that the overall effect of UVB is negative regardless of the spectral quantity or quality of available light. The effects of UVB radiation may vary over developmental stages (e.g., Häkkinen et al., 2001; Altamirano et al., 2003a; Hessen, 2003). However, variation in survival at different life history stages may have little impact at the population or community levels. For example, some amphibian species may be especially sensitive to UVB in early life stages but this may not affect them at the population level (Vonesh & De La Cruz, 2002). Regardless, in our analysis, survival under UVB radiation was lower than shielded controls in all life-history stages. Furthermore, embryos tend to grow more slowly when exposed to UVB radiation as illustrated in our analysis. Hampered growth in embryos could be due to several factors. Most importantly, the rate of cellular division is generally higher in embryos compared with other life history stages and DNA repair may be limited during rapid divisions within an embryo (Epel et al., 1999). Thus, development could be slowed by a damaged genome. Alternatively, development may be delayed by time-intensive repair

51 31 mechanisms (Epel et al., 1999). However, this result is ambiguous because of the relatively few (4) comparisons of growth in embryos. More research is needed on the effects of UVB on growth of embryos to clarify the relationship between life history stage and reduced growth due to UVB exposure. Regardless of the differences in inherent susceptibility between life history stages, embryos are generally non-motile, so behavioural avoidance of UVB in natural systems is unlikely unless oviposition occurs in a shielded environment. Therefore, embryos may be exposed to high levels of UVB during development. In contrast, motile stages and mobile species may prevent damage by behaviourally avoiding UVB radiation (Banaszak, 2003; Blaustein & Belden, 2003). Later life history stages of animals and many species of phytoplankton actively select microhabitats that may reduce exposure to UVB radiation (Häder, 1993). However, there may be a trade-off between exposure to warm sunlit areas with higher levels of PAR (optimal for photosynthesis and thermoregulation) and avoidance of areas with harmful levels of UVB (Hutchison & Duprè, 1992; Häder, 1993). Negative effects of reduced growth during early life history stages on lifetime fitness have been demonstrated in several taxa including insects (Mouer & Istock, 1980), birds (Sedinger et al., 1995) and amphibians (Semlitsch et al., 1988). These delayed life history effects are rarely explicitly incorporated into theoretical models of population and community dynamics, but are important to our understanding of how early environments and conditions affect population fluctuations in natural systems (Beckerman et al., 2002).

52 32 Previous reviews have suggested that primary producers, particularly phytoplankton, may be especially sensitive to UVB radiation (e.g., Häder, 1993; Day & Neale, 2002). In our analysis, no differences in survival were observed between trophic groups. In the survival analysis, the mean effect size in primary producers was smaller than in consumers and not different from zero. Few studies have examined survival in primary producers and the wide confidence interval may reflect the small sample size (N = 6) in our analysis. Clearly, more research on the effects of UVB on survival of primary producers is necessary to determine the importance of UVB on trophic interactions. Our growth analysis suggests that UVB radiation may affect primary producers more than consumers, although the trend was only marginally significant and disappeared when the very large effect sizes were removed from the analysis. This trend may be driven by the large negative effect of UVB on protozoans. However, more research on the effect of UVB radiation on growth in primary producers is needed to clarify this trend. In particular, more work on the effects of UVB radiation on protozoans is necessary. Of the nine comparisons in the kingdom Protozoa, seven were of dinoflagellates and eight of the nine were the work of one laboratory (Ekelund, 1990; Ekelund, 1991; Ekelund, 1993). In mesocosm experiments with plankton communities, several studies reported a larger negative effect of UVB on phytoflagellates (protozoans) compared to diatoms (Villafañe et al., 1995; Hernando et al., 2006) but in other experiments the opposite trend was observed (Wängberg et

53 33 al., 1996). Reduced growth of primary producers may lead to bottom-up control of these systems due to diminished food resources. Previous reviews highlight the variation in susceptibility to UVB between organisms (de Mora et al., 2000; Helbling & Zagarese, 2003). Our analyses demonstrate this variation in the distribution of effect sizes represented by mean and 95% confidence intervals. Effect sizes in these analyses ranged from to 5.20 (Tables 2.1 and 2.2). Although the overall effect was large and negative, individual species may be more susceptible to damage from UVB. Our random effects model allowed for a distribution of effect sizes and the estimates of pooled SD were relatively large ( and in the survival and growth analyses, respectively). Varying resistance to damage from UVB may lead to shifts in diversity or richness in both freshwater and marine phytoplankton populations, as has been observed in zooplankton communities (Marinone et al., 2006). Community composition may shift to favour a microbial web over a heterotrophic web (e.g., Mostajir et al., 1999). These shifts in community composition, diversity, or species richness in addition to the effects of UVB on dissolved carbon may alter the carbon dynamics in the oceans (Mostajir et al., 2000). The majority of experiments on community-level effects of UVB focus on one component of a natural community (i.e., phytoplankton community). Including more community components may reveal shifts in community structure that are not predicted based on sensitivity to UVB alone. For example, Bothwell et al. (1994) observed indirect positive effects of UVB radiation on algae due to a reduction in herbivory. The effects of UVB on

54 34 communities may be transient. A recent study by Wahl et al. (2004) found no difference in diversity or biomass between marine benthic communities exposed to UVB and communities shielded from UVB after 12 weeks. More long term experiments on communities are necessary to fully understand the effects of UVB radiation on diversity, richness and ecological function. Although we predicted that the effects of UVB would vary along taxonomic groupings, significant heterogeneity persisted through all taxonomic levels. Partitioning variance using the level of kingdom was impossible in the survival analysis as the vast majority of comparisons in this analysis focused on members of the kingdom Animalia (80 out of 86). Even within the kingdom Animalia, our attempts to partition variance through taxonomic structure were unsuccessful and heterogeneity persisted in each model. Similarly, heterogeneity persisted in every model in the growth analysis. This variation in effect size most likely reflects both intra- and interspecies variation in susceptibility to UVB radiation in addition to variation due to experimental conditions such as optical characteristics of water, timing of UVB exposure and dose rate. In an attempt to quantify the relationship between experimental conditions and effect size we used dose and dose-rate variables that included hours of UVB per day, total hours of UVB, days of UVB exposure, and daily dose. In the survival analysis we were also able to include total erythemal dose (erythemal dose per day summed across all exposure days). We could not use optical characteristics in our analysis as most authors do not report these types of data (e.g., extinction coefficients). We

55 35 predicted a relationship between dose and effect size, as many organisms respond to UVB with a dose-response curve (Damkaer, 1981; McNamara & Hill, 1999; Ankley et al., 2002; Browman et al., 2003; Hessen, 2003). In all cases the fit to the model was nonsignificant, suggesting that the relationship between dose-rate variables and effect size was weak in the survival analysis. In the growth analysis, the regression term was nonsignificant in all models except days of UVB exposure. Although the regression term was significant in the days of exposure regression model, the residual error term was highly significant; thus, the overall fit to the model was poor. These analyses did not detect a strong relationship between dose-rate variables and effect size. Interspecific variation in conjunction with experimental variation may obscure the dose/effect size relationship in our analyses. Broader impacts and conservation implications To our knowledge, these analyses are the first quantitative evidence of the overall negative impact of UVB radiation on aquatic organisms. Traditional reviews of the effects of UVB radiation are unable to detect the broad patterns revealed by our meta-analyses. These reviews emphasize the variation between organisms, habitats, life-history stages, and trophic levels (e.g., Siebeck et al., 1994; de Mora et al., 2000; Häder et al., 2003; Helbling & Zagarese, 2003). Our analyses captured this variation through the distribution of effect sizes, but also reveal a strong negative effect of UVB despite this variation. The dynamics of UVB exposure and resulting organismal damage is complex in natural systems. The effects of UVB in both freshwater and marine systems are modulated by many factors including seasonality of UVB dose,

56 36 total ozone concentration in the stratosphere, cloudiness, local topography, DOC, organismal behaviour and repair mechanisms. These factors may vary widely between studies, habitats, developmental stage, or species; however, these analyses emphasize the commonality of a negative effect of UVB radiation. The most striking and important result of these analyses is the consistency of the effect of UV regardless of other moderating variables within each study. These variables may have a large impact within a study, but when all the data were combined, the majority of comparisons showed a negative effect of UVB that was within the expected distribution of effect sizes. Moreover, those studies that did not fall within the expected distribution had larger (more negative) effect sizes. Our analyses highlight the importance of UVB radiation in both marine and freshwater organisms. The response variables selected for these analyses are only two of the many possible effects of UVB radiation; therefore, it is likely that these analyses underestimate the potential effects of UVB in natural systems. For example, effects such as reduced photosynthetic rates, tissue damage, and behavioural changes have been documented in many species (reviewed in Tevini, 1993; de Mora et al., 2000; Helbling & Zagarese, 2003). If UVB radiation has a negative effect on all of these variables, the overall influence of UVB could be high in these systems. Moreover, predicted increases in acidification may reduce the DOC levels in freshwater systems, resulting in higher UVB exposure in these systems (Vinebrooke et al., 2004). As a consequence of global environmental change, stressors such as UVB, chemical contaminants, drought, disease, and acidification are increasingly common in

57 37 natural systems. We did not include additional stressors in these analyses, but it is unlikely that a system would be exposed to only one stressor at a time. Environmental stressors such as UVB may interact with other environmental or biotic stressors and result in non-additive responses that are larger than predicted by each stressor individually (Vinebrooke et al., 2004). For example, UVB radiation acts synergistically with other stressors such as contaminants, disease, and extreme thermal events (Kiesecker & Blaustein, 1995; Häder et al., 2003; Pelletier et al., 2006). Alternatively, one stressor may have an antagonistic effect on other stressors, such that exposure to two stressors is less than additive (Christensen et al., 2006). At the community level, differences in susceptibility to environmental stressors may vary between organisms, leading to unforeseen interactions between stressors on the community as a whole. For example, one species may be more susceptible to chemical contaminants and less susceptible to UVB radiation, while another species is less susceptible to chemical contaminants but more susceptible to UVB radiation. Because these two stressors may be found in the same habitat, the overall effect of the stressors may be greater than predicted considering each stressor alone. Exposure to multiple stressors may shift communities towards dominance by a few hardy species (Christensen et al., 2006). Synergisms among stressors are increasingly important in the face of global environmental change and must not be ignored when considering both the effects of UVB on a single species and the effects of UVB on entire communities and systems.

58 38 Acknowledgements We thank L. Petes, E. Borer and J. Lawler for their valuable comments on earlier drafts of this manuscript. Comments from three anonymous reviewers greatly improved this manuscript. We thank R. Bancroft, N. Flynn, R. Archibald and C. Miller for assistance.

59 Table 2.1. Summary information for each comparison included in the survival analysis. M = marine, FW = freshwater, FD = field, LB = lab, P = primary producer, C = consumer, E = embryo, L = larvae, A = adult, N = naupli, CP = copepodid, S = spore, J = juvenile. Species Habitat Venue Trophic Level Lifehistory stage Effect size (d) Variance of d Source Acartia ornorii M LB C E * Lacuna & Uye 2001 Acropora palmata M FD C L -0.57* Wellington & Fitt 2003 Agaricia agaricites M FD C L -1.94* 0.74 Gleason & Wellington 1995 Ambystoma gracile FW FD C E -4.58* 1.81 Blaustein et al Ambystoma maculatum Boeckella brevicaudata FW FD C E Starnes et al FW FD C A Zagarese et al Boeckella gibbosa FW FD C A Zagarese et al Boeckella gracilipes FW FD C A -1.47* 0.51 Zagarese et al Boeckella gracilipes FW FD C N Cabrera et al Boeckella gracilipes FW FD C CP Cabrera et al Boeckella gracilipes FW FD C A Cabrera et al

60 Boeckella gracilipes FW FD C A -2.25* 1.09 de los Ríos & Soto 2005 Brachydanio rerio FW LB C E Charron et al Brachydanio rerio FW LB C L * 9.82 Charron et al Bufo americanus FW LB C L Grant & Licht 1995 Bufo boreas FW LB C J -3.85* 0.57 Blaustein et al Bufo boreas FW FD C E Corn 1998 Bufo boreas FW FD C E Corn 1998 Bufo bufo FW FD C E Häkkinen et al Bufo bufo FW FD C L -8.50* 5.01 Häkkinen et al Calanus finmarchicus M FD C E Browman et al Cancer magister M LB C L * 5.54 Damkaer et al Cancer oregonesis M LB C L Damkaer et al Chaetoceros brevis M LB P A van de Poll et al Chaetoceros brevis M FD P A van de Poll et al Chondrus crispus M LB P S -0.79* 0.72 Roleda et al

61 Coregonus albula M LB C L Häkkinen et al Coregonus lavaretus M LB C L Häkkinen et al Coregonus lavaretus M LB C L Ylönen & Karjalainen 2004 Corella inflata M FD C J Bingham & Reitzel 2000 Daphnia magna FW LB C A -1.34* 0.41 Borgeraas & Hessen 2000 Daphnia pulex FW FD C A -3.48* 1.68 de los Ríos & Soto 2005 Daphnia pulicaria FW LB C A Williamson et al Daphnia pulicaria FW FD C A -3.32* 2.38 Zagarese et al Esox lucius FW FD C L Häkkinen & Oikari 2004 Gadus morhua M FD C E -2.44* 0.50 Beland et al Gadus morhua M FD C E * Browman et al Heterocapsa triquetra M LB P A Wängberg et al Homarus americanus M LB C L Rodriguez et al Hyla cadaverina FW FD C E -2.42* 0.43 Anzalone et al Hyla chrysoscelis FW FD C E Starnes et al

62 Hyla regilla FW FD C E Anzalone et al Keratella cochlearis M FD C A -1.55* 0.43 Vinebrooke & Leavitt 1999 Keratella quadrata M FD C A -2.67* 0.63 Vinebrooke & Leavitt 1999 Lepadella sp. M FD C A -1.24* 0.40 Vinebrooke & Leavitt 1999 Lepomis macrochirus FW FD C E -1.30* 0.61 Gutiérrez-Rodríguez & Williamson 1999 Lepomis macrochirus FW FD C E -1.14* 0.58 Gutiérrez-Rodríguez & Williamson 1999 Litoria aurea FW FD C E van de Mortel & Buttemer 1996 Litoria aurea FW FD C E van de Mortel & Buttemer 1996 Litoria dentata FW FD C E van de Mortel & Buttemer 1996 Litoria dentata FW FD C E van de Mortel & Buttemer 1996 Litoria peronii FW FD C E van de Mortel & Buttemer 1996 Litoria peronii FW FD C E van de Mortel & Buttemer 1996 Mastocarpus stellatus M LB P S -2.02* 1.01 Roleda et al Montastraea annularis M FD C L -8.24* 0.54 Wellington & Fitt

63 Montastraea franksi M FD C L -2.42* 0.10 Wellington & Fitt 2003 Notholca sp. M FD C A Vinebrooke & Leavitt 1999 Oncorhynchus apache FW LB C L 5.20* 2.92 Little & Fabacher 1994 Oncorhynchus clarki henshawi FW LB C L -4.40* 2.28 Little & Fabacher 1994 Pandalus hypsinotus M LB C L -1.54* 0.52 Damkaer et al Pandalus platyceros M LB C L -0.58* 0.42 Damkaer et al Psuedacris crucifer FW LB C L -4.18* 1.59 Baud & Beck 2005 Psuedacris triseriata FW FD C E Starnes et al Rana arvalis FW FD C E -2.94* 1.04 Häkkinen et al Rana arvalis FW FD C L Häkkinen et al Rana arvalis FW FD C E Pahkala et al Rana arvalis FW LB C E Pahkala et al Rana blairi FW FD C E Smith et al Rana cascadae FW FD C L -1.13* 0.19 Belden et al

64 Rana cascadae FW LB C L Hatch & Blaustein 2002 Rana clamitans FW FD C L -9.32* Tietge et al Rana luteiventris FW FD C E Blaustein et al Rana pipiens FW LB C L Ankley et al Rana pipiens FW FD C L -6.07* 5.60 Ankley et al Rana pipiens FW FD C L * Tietge et al Rana pretiosa FW FD C E Blaustein et al Rana septentrionalis FW FD C L * Tietge et al Rana sylvatica FW LB C L Grant & Licht 1995 Rana temporaria FW FD C E Häkkinen et al Rana temporaria FW FD C L Häkkinen et al Rana temporaria FW LB C E Pahkala et al Rana temporaria FW FD C E Merilä et al Rana temporaria FW FD C E Pahkala et al Sinocalanus tenellus M LB C E * Lacuna & Uye

65 Taricha torosa FW FD C E -2.97* 0.25 Anzalone et al Thalassiosira sp M FD P A Hernando et al Thysanoessa raschii M LB C L -1.45* 0.51 Damkaer et al * Significant effect reported in original article 45

66 Table 2.2. Summary information for each comparison included in the growth analysis. M = marine, FW = freshwater, FD = field, LB = lab, P = primary producer, C = consumer, E = embryo, L = larvae, A = adult, SP = sporophyte, GM = gametophyte, J = juvenile. Species Habitat Venue Trophic Level Lifehistory stage Effect size (d) Variance of d Source Ambystoma gracile FW LB C L -4.07* 0.88 Belden & Blaustein 2002a Ambystoma macrodactylum Ambystoma macrodactylum Ambystoma macrodactylum FW LB C L -1.29* 0.35 Belden & Blaustein 2002b FW LB C L -0.55* 0.05 Belden & Blaustein 2002c FW LB C L -0.42* 0.04 Belden & Blaustein 2002c Amphistegina gibbosa M LB C A Williams & Hallock 2004 Bufo americanus FW LB C L Grant & Licht 1995 Chondrus crispus M LB P GM Roleda et al. 2004a Coregonus albula M LB C L Häkkinen et al Coregonus lavaretus M LB C L Ylönen & Karjalainen

67 Coregonus lavaretus M LB C L Häkkinen et al Cylindrotheca closterium M LB P A -2.47* 0.88 Rijstenbil 2003 Ditylum brightwellii M LB P A -0.88* 0.73 Ekelund 1990 Ditylum brightwellii M LB P A 0.62* 0.70 Ekelund 1990 Enteromorpha intestinalis Enteromorpha intestinalis M LB P SP * 6.72 Cordi et al M LB P GM -0.91* 0.44 Cordi et al Euglena gracilis FW LB P A * 3.27 Ekelund 1993 Fucus gardneri M LB P E * Henry & Van Alstyne 2004 Fucus gardneri M LB P J Henry & Van Alstyne 2004 Fucus serratus M LB P E -3.48* 1.68 Altamirano et al. 2003a Fucus serratus M LB P GM Altamirano et al. 2003b Fucus spiralis M LB P GM -4.55* 2.40 Altamirano et al. 2003b Fucus vesiculosus M LB P GM * 9.00 Altamirano et al. 2003b Gyrodinium aureolum M LB P A -9.08* 7.54 Ekelund

68 Gyrodinium aureolum M LB P A -5.68* 3.36 Ekelund 1990 Gyrodinium aureolum M LB P A -1.79* 0.93 Ekelund 1991 Heterocapsa triquetra M LB P A -2.53* 1.20 Ekelund 1991 Hyla versicolor FW LB C L Grant & Licht 1995 Laminaria ochroleuca M LB P SP Roleda et al. 2004b Mastocarpus stellatus M LB P GM Roleda et al. 2004a Nannochloropsis gaditana Phaeodactylum tricornutum Phaeodactylum tricornutum M LB P A -3.81* 2.81 Sobrino et al M LB P A Ekelund 1990 M FD P A 1.63* 0.67 Behrenfeld et al Prorocentrum minimum M LB P A * Ekelund 1990 Prorocentrum minimum M LB P A -7.36* 5.18 Ekelund 1991 Prorocentrum minimum M LB P A -3.04* 1.44 Ekelund 1990 Pseudonitzschia seriata M FD P A -6.06* 2.24 Nilawati et al Rana arvalis FW FD C E Pahkala et al. 2001a 48

69 Rana pipiens FW LB C L -1.58* 0.66 Crump et al Rana pipiens FW LB C L Ankley et al Rana sylvatica FW LB C L 1.83* 0.71 Grant & Licht 1995 Rana temporaria FW LB C E Pahkala et al. 2001b Selenastrum capricornutum Selenastrum capricornutum FW LB P A -3.42* 0.55 West et al FW LB P A -0.75* 0.36 West et al Thalassiosira sp. M FD P A -3.78* 1.86 Hernando et al Ulva expansa M FD P A Grobe & Murphy 1998 Ulva rigida M FD P A -1.21* 1.18 Altamirano et al * Significant effect reported in original article 49

70 Table 2.3.Heterogeneity statistics for each model in the survival analysis. Separate analyses were conducted to compare similarity in effect size between each group. Statistical model df Q Probability Full model (no structure) < Habitat type Between groups Within groups < Experimental venue Between groups Within groups < Trophic group Between groups Within groups < Developmental stage Between groups Within groups <

71 Table 2.4. Results from variance partitioning using taxonomic group in the survival analysis. Variance between mean effect size estimates for each group was compared with other members of the preceding hierarchical level only (i.e., only classes within a phylum were compared to each other). Statistical model df Q Probability Representative groups Phylum (within Animalia) Between groups Within groups < Class (within Arthropoda) Between groups Within groups Class (within Chordata) Between groups Within groups < Family (within Anura) Between groups Within groups Arthropoda (20), Chordata (53) Cnidaria (4), Rotifera (4) Branchiopoda (4), Copepoda (9) Malacostraca (6) Actinopterygii (12), Amphibia (40) Bufonidae (6), Hylidae (11) Ranidae (20) 51

72 Table 2.5. Relationship between mean effect size estimates and dose-rate variables in the survival analysis. Parameter Slope Source df Q P Hours of UVB per day Regression Residual <0.001 Total hours of UVB Regression Residual <0.001 Days of exposure Regression Residual <0.001 Dose (DNA) Regression Residual Dose (Erythemal) Regression Residual Cumulative dose Regression (Erythemal) Residual

73 Table 2.6. Heterogeneity statistics for each model in the growth analysis. Separate analyses were conducted to compare similarity in effect size between each group. Statistical model df Q Probability Full model (no structure) < Habitat type Between groups Within groups < Experimental venue Between groups Within groups < Trophic group Between groups Within groups < Developmental stage Between groups Within groups <

74 Table 2.7. Results from variance partitioning using taxonomic group in the growth analysis. Variance between mean effect size estimates for each group was compared with other members of the preceding hierarchical level only (i.e., only classes within a phylum were compared to each other). 54 Statistical model df Q P Representative groups Kingdom Between groups Within groups < Phylum (within Plantae) Between groups Within groups Animalia (14) Chromista (11) Plantae (12) Protozoa (9) Bacillariophyta (4) Chlorophyta (6) Order (within Amphibia) Between groups Within groups Anura (7) Urodela (4)

75 Table 2.8. Relationship between mean effect size estimates and dose-rate variables in the growth analysis. Parameter Slope Source df Q P 55 Hours of UVB per day Regression Residual < Total hours of UVB Regression Residual < Days of exposure Regression Residual < Dose (DNA) Regression Residual Dose (Erythemal) Regression Residual

76 56 a b Figure 2.1.Normal quantile plots of effect size for survival (a) and growth (b) analyses. Dashed lines indicate 95% confidence intervals. Each point represents the effect size calculated for one comparison in the analysis. The distribution of effect sizes is linear and within the bounds of the 95% confidence interval in both plots, suggesting the data are normally distributed.

77 Figure 2.2.The effect of UVB radiation on survival. The mean and 95% confidence interval is shown for each analysis. The number of comparisons used to calculate each mean is shown in parentheses. Confidence intervals that overlap the dashed line at zero are not significantly different from zero. FM = full model, FW = freshwater, M = marine, FD = field, LB = laboratory, C = consumer, P = primary producer, E = embryo, L = larva, A = adult. 57

78 Figure 2.3.Effect of UVB radiation on growth. The mean and 95% confidence interval is shown for each analysis. The number of comparisons used to calculate each mean is shown in parentheses. All means are significantly different from zero (95% confidence intervals do not overlap with the dashed line at zero). FM = full model, FW = freshwater, M = marine, FD = field, LB = laboratory, C = consumer, P = primary producer, E = embryo, L = larva, A = adult. 58

79 Figure 2.4. Effect of UVB radiation on growth in each kingdom. The mean and 95% confidence interval are shown for each kingdom. The means are all significantly different from zero as none of the 95% confidence intervals overlap zero. UVB radiation has a significantly larger negative effect on members of the kingdom Protozoa (asterisk). AN = Animalia, CH = Chromista, PL = Plantae, PR = Protozoa. 59

80 Figure 2.5. Weighted histogram of effect sizes for all comparisons in the survival analysis (a) and the growth analysis (b). The height of each bar indicates the combined weight of effect sizes in each class. The distribution is left-skewed due to several extreme negative values with low weight. 60

81 Figure 2.6. Weighted histogram of effect sizes after removal of effect sizes greater than 1 standard deviation from zero. The height of each bar indicates the combined weight of effect sizes in each class. The distribution of weighted effect sizes for the survival analysis (a) and the growth analysis (b) are normal and show no evidence of publication bias. 61

82 62 Chapter 3 A meta-analysis of the effects of ultraviolet B radiation and other stressors on survival in amphibians Betsy A. Bancroft, Nick J. Baker, and Andrew R. Blaustein Submitted to Global Change Biology

83 63 Abstract Human alterations to natural systems have resulted in a loss of biological diversity in many species and locations around the world, including amphibian populations. Hypotheses for the decline of amphibian populations include habitat loss, introduction of exotic species, pathogens, pollution of habitats, and global environmental changes such as increases in ultraviolet-b (UVB) radiation reaching the earth s surface. UVB negatively affects many amphibian species and can cause mortality and sublethal effects, such as reduced growth and increased susceptibility to disease. In addition, UVB radiation can act synergistically with natural and anthropogenic stressors in amphibian habitats, resulting in greater than additive effects. However, the effect of UVB varies widely among species, and can also vary within populations of the same species or at different life history stages. This variation, in combination with a lack of significant effects of UVB in much of the published literature, has resulted in controversy regarding the importance of UVB as a stressor for amphibians. We used meta-analysis techniques to explore the overall effects of UVB radiation on survival in amphibians. We also used recently developed factorial meta-analytic techniques to quantify potential interactions between UVB radiation and other stressors on amphibians. UVB radiation reduced survival of amphibians by 1.9-fold compared to shielded controls. Days of exposure, life-history stage, and taxonomic order accounted for the majority of variation in effect size in multiple linear regression. Larvae are more susceptible to damage from UVB radiation compared to embryos, and caudates are more susceptible compared to anurans. Furthermore, UVB radiation

84 64 interacts synergistically with other environmental stressors and results in greater than additive effects on survival when two stressors are present. Our results suggest that UVB radiation is an important stressor in amphibians, particularly in light of potential synergisms between UVB and other stressors in amphibian habitats. Introduction Anthropogenic changes to the environment have altered the abiotic and biotic habitat of many organisms. These changes include the addition of contaminants (Fleeger et al., 2003), introduction of exotic organisms (Sakai et al., 2001), alteration of flood (Gergel et al., 2005) and fire regimes (Allen et al., 2002), and increases in mean temperatures (IPCC, 2007). In addition, reduction in the stratospheric ozone layer has resulted in increases in ultraviolet-b (UVB) radiation reaching the earth s surface (Kerr & McElroy, 1993; Madronich, 1993; Madronich et al., 1998; Solomon, 1999). UVB negatively affects a wide variety of freshwater and marine organisms (Bancroft et al., 2007). The effects of UVB range from increased mortality to sublethal effects on growth, development, photosynthesis and immunity (Tevini, 1993; Caldwell et al., 1998). These effects may scale up to the population or community level, causing changes in community structure and function (Bothwell et al., 1994; Mostajir et al., 1999; Marinone et al., 2006; but see Wahl et al., 2004). Both abiotic and biotic alterations of habitat have negatively affected many organisms, including amphibians. Amphibians are of particular conservation concern, as amphibian populations are declining more rapidly than either birds or mammals

85 65 (Stuart et al., 2004) with perhaps as many as 122 species becoming extinct since 1980 (Mendelson et al., 2006). Many factors appear to be contributing to amphibian population declines. These include habitat loss, introduced species, pathogens, contaminants, climate change, and increases in UVB radiation (Blaustein & Kiesecker, 2002; Collins & Storfer, 2003). The diversity of locations where amphibian populations have declined prompted consideration of atmospheric factors e.g., increased terrestrial UVB irradiance associated with stratospheric ozone depletion (Blaustein et al., 1994). The effects of UVB radiation on amphibians include increased mortality, reduced growth, developmental abnormalities, increased susceptibility to disease and behavioural changes (reviewed in Blaustein et al., 1998; Blaustein & Kiesecker, 2002). In addition, the magnitude of the effect may vary between species or between different populations or life history stages of the same species (e.g., Blaustein et al., 1998). For example, survival of moor frog (Rana arvalis) embryos was higher when UVB radiation was filtered out compared with embryos exposed to UVB radiation (Häkkinen et al., 2001). However, survival of larval moor frogs was not affected by UVB exposure (Häkkinen et al., 2001). In contrast, survival of embryonic common toads (Bufo bufo) was not affected when exposed to UVB, but survival was lower in larvae exposed to UVB compared with those shielded from UVB (Häkkinen et al., 2001). Numerous factors may influence the results and interpretation of experimental studies reporting the effects of UVB radiation on living organisms (Bancroft et al., 2007). These factors may be aspects of the environment (e.g., altitude, climate,

86 66 latitude) or even the methods used by the investigators. For example, venue (laboratory or field) and the duration of an experiment often vary between studies, which may make it difficult to compare results between studies. Venue may be particularly important, as the quality and quantity of UVA ( nm) and photosynthetically active radiation (PAR) in laboratory experiments rarely approximates conditions in the field (e.g., Ankley et al., 2000). These wavelengths are necessary for effective DNA repair after exposure to UVB radiation (Sancar & Sancar, 1988). Lack of efficient repair or exposure to high doses of UVB radiation that overwhelm repair mechanisms may be responsible for the observed negative effects of UVB in amphibians. Many studies suggest that UVB may interact synergistically with other stressors such as contaminants, climatic factors or pathogens (Blaustein et al., 2001, 2003). For example, western toad (Bufo boreas) embryos are susceptible to a complex interaction between UVB radiation, a pathogenic water mould (Saprolegnia sp.), and changes in precipitation (Kiesecker et al., 2001). Thus, mortality in western toad embryos increases when they are infected with Saprolegnia in the presence of higher UVB radiation which occurs during years of lower precipitation when water levels are low and the UVB shielding property of the water is diminished. Climate cycles such as El Niño and La Niña, with associated changes in precipitation, affect this dynamic (Kiesecker et al., 2001). Several studies have reported no effects of UVB alone or in conjunction with other stressors on amphibians (e.g., Blaustein et al., 1996; Starnes et al., 2000; Merilä

87 67 et al., 2000; Pahkala et al., 2001). The lack of a UVB effect on amphibians reported by many studies has led to controversy regarding the importance of UVB as an environmental stressor in habitats where amphibians live. This controversy stems from using formal or informal vote count summaries of the literature, where the numbers of studies reporting significant effects are summed and a conclusion about the overall importance of a factor is estimated (e.g., Licht, 2003). However, this method suffers from low statistical power and is biased towards finding no effect (Rosenberg et al., 2000). In contrast, meta-analytic techniques are designed to avoid these issues. Meta-analysis techniques are less biased and have been used to explore the effects of UVB radiation and other global environmental factors (e.g., Root et al., 2003; Parmesan & Yohe, 2003; Bancroft et al., 2007). Furthermore, recent advances in meta-analytic techniques now allow for assessing the potential interaction among factors in fully factorial experiments (Gurevitch et al., 2000; Hawkes & Sullivan, 2001). We used meta-analytic techniques to quantify the overall effect of UVB radiation on survival of amphibians. We then used multiple linear regression to explore the sources of variation in estimates of effect size and to identify important predictors of effect size in published studies. Finally, we used factorial meta-analysis to quantify the interaction between UVB and additional stressors. Our results suggest that UVB is indeed an important stressor for amphibians, as it is for many other aquatic organisms (Bancroft et al., 2007).

88 68 Materials and Methods Effects of UVB on survival We searched six electronic databases (BIOSIS, Web of Science, Aquatic Sciences and Fisheries Abstracts, Fish and Fisheries Worldwide, Wildlife and Ecology Studies Worldwide, and Biological and Agricultural Index) to identify papers used in these analyses. We searched using all possible combinations of the terms ultraviolet, UV, UVB with survival and amphibian, frog, toad, salamander or newt and identified 41 peer-reviewed articles that provided a measure of amphibian survival after exposure to ambient levels of UVB. We limited our search to include only experimental manipulations of UVB, using the standard technique of applying plastic filters that differentially transmitted or filtered UVB radiation. We included all study locations and species within each article. When multiple sampling dates were reported, we selected the final sampling day. If multiple doses of UVB radiation were used, we randomly selected only one dose from each study. One half of these articles were used in a previous analysis (Bancroft et al., 2007). The 41 articles generated 89 comparisons and included two of the three amphibian orders (anura and caudata), eight genera, and 32 individual species (Table 3.1). We extracted data from figures using TechDig 2.0 (Jones, 1998). We calculated the log-response ratio (lnr) as a measure of effect size for each study. The log-response ratio is calculated as the natural log of the ratio of survival with and without UVB radiation within each comparison (Hedges et al., 1999). We selected the lnr because of the clear biological interpretation (proportional survival of

89 69 experimental organisms compared to controls) and good statistical properties (Osenberg et al., 1997; Shurin et al., 2002). We calculated the grand mean effect size for the full model and generated bias-corrected bootstrapped 95% confidence intervals (Adams et al., 1997). We calculated the grand mean effect size using unweighted effect sizes because approximately half of the comparisons did not include a measure of variance. Effect size calculations and summary analyses were conducted in MetaWin 2.0 (Rosenberg et al., 2000). Using unweighted effect size estimates decreases the probability of detecting differences among groups; however, restricting our analyses to comparisons with estimates of precision would drastically reduce our sample size and could introduce bias into our analyses (Englund et al., 1999). We explored the variables contributing to residual variance in our model using multiple linear regression (Borer et al., 2005). The variables examined included biological, geographical and methodological factors reported in each study (Table 3.2). However, some studies did not include enough information to estimate study duration, days of UVB exposure, or altitude of study location. Therefore we conducted separate analyses using all variables, effectively excluding the comparisons without all variables reported, and then on the subset of variables reported by every study. We used backwards selection to identify the best fit model. Multiple linear regressions were conducted in JMP V.6 (SAS Institute Incorporated, Cary, NC, USA). We generated mean effect sizes and bias-corrected bootstrapped 95% confidence intervals (CIs) for groups with a significant term in our regression model. When fewer than 10 comparisons were available within a group, we used parametric 95% CIs as a more

90 70 conservative estimate of the CI. Mean effect sizes were considered significantly different from zero if the 95% CI did not overlap with zero. Effects of UVB and additional stressors on survival To explore the possible interaction between UVB and additional environmental stressors, we used factorial meta-analysis (Gurevitch et al., 2000). This type of metaanalysis allowed the examination of main effects (i.e., UVB and the other stressor alone), plus the interaction between the two stressors. However, only studies that were originally factorial in design were used (Gurevitch et al., 2000; Hawkes & Sullivan, 2001). We conducted a separate search in the same six databases (see above) for studies testing the effects of UVB and an additional stressor on survival alone and in combination (Table 3.3). We used the log-response ratio as our effect size metric. We calculated the main effects of each stressor and the interaction between the stressors, in addition to the individual effect sizes for each treatment (Figure 3.1). A negative value of the main effects indicates that the stressor had a negative effect on survival. A negative value of the interaction term indicates a synergistic (more than additive) effect of the two stressors together. Effect size estimates were considered significantly different from each other if the 95% CIs did not overlap. Furthermore, effect size estimates were considered significantly different from zero if the 95% CIs did not overlap with zero. Two comparisons were removed from the analysis due to mathematical incompatibility (zero survival in all treatments except controls).

91 71 Results Effects of UVB on survival Exposure to UVB resulted in a 1.85-fold decrease in survival compared to controls (Figure 3.2a). When all explanatory variables were included in the regression analysis, days of UVB exposure, life history stage and taxonomic order accounted for 41.4% of the variation in survival (Table 3.4). When all comparisons were included in the analysis, and therefore fewer explanatory variables, life history stage and taxonomic order explained 18.2 % of the variation in survival after exposure to UVB (Table 3.4). Surprisingly, the correlation coefficient for days of UVB exposure was positive (0.012), suggesting that amphibian survival was higher with longer exposure times. Larvae were more sensitive to UVB than embryos (Figure 3.2b), and caudates were more susceptible to UVB than anurans (Figure 3.2c). Nearly one third (28%) of the comparisons in these analyses were the work of A.R. Blaustein and colleagues. To explore the possibility of bias, we removed all comparisons generated by this laboratory and re-ran the analyses. The final regression models were qualitatively the same without studies conducted by the Blaustein laboratory group (Table 3.5). Without the Blaustein laboratory comparisons in the analysis, UVB reduced survival by 1.9-fold. The only change is a shift of the mean effect size for metamorphic individuals from negative (-0.22, 95% CI: to ) to positive (0.08, 95% CI: to ). Effect of UVB and additional stressors on survival

92 72 In the factorial analysis, the main effects (UVB and additional stressors) were different from zero, but not different from each other (Figure 3.3a). There was evidence of a synergistic (more than additive) effect of UVB and additional stressors, as the interaction effect size was negative and different from zero (Figure 3.3a). UVB alone had the smallest effect size estimate but was different from zero (Figure 3.3b). The combination of an additional stressor and UVB radiation had the largest effect size estimate and was also different from zero (Figure 3.3b). Discussion UVB radiation alone and in combination with other factors is an important stressor in natural systems. The interaction between stressors can lead to ecological surprises even when one stressor is not considered to be a major stressor in the system (Christensen et al., 2004). Effect of UVB alone Exposure to UVB radiation reduced survival of amphibians by approximately 1.9-fold. This result was surprising because only 32 of the original 89 comparisons (35%) reported a significant reduction in survivorship under UVB radiation. This analysis clearly illustrates the power of meta-analysis to reveal patterns in published literature that are otherwise obscured by low statistical power. The majority of the variation in effect size was explained by life history stage, taxonomic order and the number of days of exposure.

93 73 The average effect size was much smaller for embryos than larvae, suggesting that UVB radiation has a larger effect on survival of larvae. Only four studies on the effects of UVB on metamorphic amphibians were included. Clearly, more research is necessary on the effects of UVB at this important life history stage. Embryos may have been less susceptible to UVB radiation because they were exposed to UVB for fewer days on average (embryos, 12.5 days; larvae, 34.5 days), most likely due to the shorter period spent in this life history stage. In addition, embryos may be better protected from UVB by the jelly envelope, pigmentation, or cellular repair mechanisms such as photolyase (Epel et al., 1999; Blaustein & Belden, 2003). Larvae may be less defended from UVB radiation because they have the potential to behaviourally avoid areas with high UVB levels in natural systems. The experimental design of the comparisons in this analysis did not allow larvae to seek refuge from UVB. Larvae of some species appear to avoid regions with high UVB exposure (Nagl & Hofer, 1997; van de Mortel & Buttemer, 1998; Garcia et al., 2004). In contrast, larvae of other species do not avoid UVB and are frequently observed in very shallow water (O Hara, 1981; Wollmuth et al., 1987; van de Mortel & Buttemer, 1998; Belden et al., 2000; Belden et al., 2003). Larger effects of UVB on survival of larvae may have important implications for population persistence. A recent demographic model suggests that reduction in post-embryonic survival results in decreases in equilibrium density in the adult portion of the population (Vonesh & De la Cruz, 2002). Therefore, reduction in larval survival due to UVB may result in

94 74 reductions in the adult population over time, particularly in habitats with high transmission of UVB and in species that do not avoid UVB radiation. The average effect size estimate was much larger for caudates compared to anurans, suggesting that UVB has a much larger effect on survival in salamanders and newts compared to frogs and toads. This result is in accordance with previous work on relative photolyase activity in embryos. Photolyases are a group of enzymes responsible for the majority of DNA repair after exposure to UVB radiation (Sancar & Sancar, 1988). In general, salamander and newt species have much lower levels of photolyase compared to frogs and toads (Blaustein et al., 1994; Smith et al., 2002). These differences in photolyase levels and the observed difference in effect size between caudates and anurans may reflect the differences in expected exposure to UVB in natural systems, particularly at the egg stage (Blaustein et al., 1994). For example, oviposition occurred at deeper depths on average for two salamander species compared to five frog species (Smith et al., 2002). However, this relationship is not always observed and some caudates lay their eggs in shallow water. Oviposition depth varies between species and even between individuals of the same population (Palen et al., 2005). Mobile larvae, however, may be exposed to varying levels of UVB due to habitat use. If larvae do not avoid UVB radiation, other factors influencing habitat use may lead to increased UVB exposure. Temperature is one important predictor of habitat use in amphibians, including caudates (Lucas & Reynolds, 1967; Keen & Schroeder, 1975; Dupré & Petranka, 1985; Hutchison & Dupré, 1992). Temperature preferences vary by species and life-history stage, but

95 75 many amphibians select warmer temperatures where growth and development are maximized (Hutchison & Dupré, 1992). Seeking warm, shallow water for thermoregulation also may expose larvae to high levels of UVB during development. Thus, caudate larvae may be exposed to similar UVB levels as anuran larvae, and these UVB levels may result in larger effects on survival in caudates. We expected the relationship between days of exposure and effect size to be negative, as the effects of UVB on an organism are closely related to total dose of UVB (e.g., Ankley et al. 2002). However, we observed the opposite trend in this analysis. It is possible that organisms surviving the first day of UVB exposure could develop a UV-hardening response, similar to the temperature hardening response seen in many organisms (Hutchison & Maness, 1979; Nobel, 1982; Lee et al., 1987). UV hardening in amphibians may involve the use of melanin as a sunscreen pigment (Jablonski, 1998; Blaustein & Belden, 2003) or rely on up-regulation of cellular mechanisms such as photolyases, heat shock proteins or antioxidant enzymes (e.g., superoxide dismutase) (Feder, 1999; Lesser et al., 2001; Blaustein & Belden, 2003). Alternatively, a priori estimates of UV tolerance could result in longer exposure times for species or stages with higher tolerance for UVB. Interestingly, a negative correlation between days of UVB exposure and effect size was observed for embryos (r = -0.38, p = 0.013), while a positive correlation was observed for larvae (r = 0.57, p = 0.006). Thus, the relationship between effect size and days of UVB exposure is in the expected direction for embryos but not larvae. This difference in the relationship between exposure days and effect size suggests differences may exist in physiological

96 76 tolerance for UVB exposure between life history stages. Embryos may be more likely to accrue damage from longer term exposure, while larvae may be more likely to exhibit UV hardening or acclimation. More research on the physiological mechanisms of UV tolerance at the larval life history stage is necessary. Effect of UVB and additional stressors on survival Our results suggest that the combination of UVB radiation and an additional stressor leads to a synergistic interaction and greater than additive mortality in amphibians. These additional stressors were diverse and ranged from pathogens to agricultural contaminants and low ph. The effect of these stressors was large and significantly different from zero when amphibians were exposed to the stressor alone. However, amphibians in natural systems are frequently exposed to more than one stressor and these stressors commonly interact synergistically (Sih et al., 2004). The finding that UVB radiation interacts with other stressors is particularly important, as many amphibian habitats in temperate regions are exposed to some degree of UVB (Diamond et al., 2005). Levels of UVB radiation vary widely both between years and habitats and at smaller spatial and temporal scales. Thus, the effects of these stressors in natural populations could vary on similar spatial and temporal scales. The potential for interaction with other stressors suggests that UVB is an important factor even in species that are relatively resistant to UVB radiation. For example, Pacific treefrogs (Pseudacris regilla) are generally considered resistant to damage from UVB radiation. However, when Pacific treefrog embryos were exposed to UVB and nitrate, survival was drastically reduced (Hatch & Blaustein, 2003).

97 77 Experiments on the same species found no interaction between UVB and a pathogen, Saprolegnia ferax (Kiesecker & Blaustein, 1995). Thus, the strength of the interaction may vary within a species depending on the stressors involved. Common frogs (Rana temporaria) are also considered relatively resistant to damage from UVB, but when both UVB and low ph were present, survival was reduced (Pahkala et al., 2002). The results of this analysis support the observation that amphibian population declines are likely due to multiple causes that may interact (Blaustein & Kiesecker, 2002). The main effect of UVB was smaller in the factorial analysis compared to the analysis of the effects of UVB alone. Fewer comparisons were included in the factorial analysis, as only fully factorial studies were included. However, none of the included studies in the factorial analysis were on caudates. In the analysis of UVB alone, UVB had a much larger negative effect on caudates compared to anurans. Furthermore, only four of the nineteen comparisons in the factorial analysis were on larvae. The rest of the comparisons were on embryos, the least susceptible life-history stage. Thus, the factorial analysis may underestimate the strength of the interaction between UVB and additional stressors. More research on caudates and later lifehistory stages is necessary to understand the potential for non-additive effects in amphibians exposed to UVB radiation. Conclusions These analyses suggest that UVB radiation is an important stressor for amphibians. A negative effect of UVB radiation on survival was detected despite the apparent lack of significant UVB effects in much of the literature. Moreover, UVB

98 78 radiation interacts synergistically with many common stressors in natural systems. Future work should focus on under-represented life-history stages and should explore the potential for interactions among stressors in natural systems. These types of data are necessary for including stressors in demographic models of amphibian populations. Understanding the effects of various stressors on multiple species of amphibians at multiple life-history stages is vital to clarifying causes of amphibian population declines. Acknowledgements We would like to thank E. Seabloom for statistical advice and E. Borer for helpful suggestions on an earlier draft of this manuscript. We thank R. Bancroft, J. Conway, K. Hill, and P. Cicero for assistance.

99 Table 3.1. Amphibian species, experimental variables and biological variables used in multiple linear regression analysis of the effect of UVB alone on survival. Study duration was defined as short for experiments < 7 days, or long for experiments > 8 days in duration. Species Latitude Study Days Elevation Venue Life history Taxonomic Reference duration UVB (m) stage order Ambystoma macrodactylum Ambystoma macrodactylum 44 Long Laboratory Larva Caudata Belden et al Long Laboratory Larva Caudata Belden et al Ambystoma gracile 44 Long Field Embryo Caudata Blaustein et al Ambystoma macrodactylum Ambystoma maculatum Ambystoma maculatum Ambystoma maculatum 44 Long Field Embryo Caudata Blaustein et al Long 17 Laboratory Embryo Caudata Lesser et al Short Laboratory Embryo Caudata Calfee et al Long Laboratory Larva Caudata Calfee et al

100 Ambystoma maculatum Ambystoma talpoideum 35 Long Field Embryo Caudata Starnes et al Short 7 14 Laboratory Larvae Caudata Calfee et al Bufo americanus 43.4 Long 63 Laboratory Larvae Anura Grant & Licht 1995 Bufo americanus 43.4 Long 14 Laboratory Metamorph Anura Grant & Licht 1995 Bufo boreas Field Embryo Anura Kiesecker & Blaustein 1995 Bufo boreas Field Embryo Anura Kiesecker & Blaustein 1995 Bufo boreas Field Embryo Anura Blaustein et al Bufo boreas Field Embryo Anura Blaustein et al Bufo boreas 40.3 Long Field Embryo Anura Corn 1998 Bufo boreas 40.3 Long Field Embryo Anura Corn

101 Bufo boreas 44 Short Laboratory Metamorph Anura Blaustein et al Bufo boreas 39 Long Laboratory Larva Anura Little et al Bufo boreas 44 Long Field Embryo Anura Kiesecker et al Bufo boreas 44 Long Laboratory Larvae Anura Worrest & Kimeldorf 1976 Bufo boreas 44 Long Laboratory Larvae Anura Worrest & Kimeldorf 1975 Bufo bufo 62 Long Field Embryo Anura Häkkinen et al Bufo bufo 62 Long Field Larva Anura Häkkinen et al Bufo bufo Field Embryo Anura Lizana & Pedraza 1998 Bufo calamita Field Embryo Anura Lizana & Pedraza 1998 Bufo woodhousii 39 Long Laboratory Larva Anura Little et al

102 Hyla cadaverina 34 Long Field Embryo Anura Anzalone et al Hyla chrysoscelis 35 Short Field Embryo Anura Starnes et al Hyla chrysoscelis 35 Short Field Embryo Anura Bruner et al Hyla (Pseudacris) regilla Hyla (Pseudacris) regilla Hyla (Pseudacris) regilla Hyla (Pseudacris) regilla Hyla (Pseudacris) regilla Hyla (Pseudacris) regilla Field Embryo Anura Kiesecker & Blaustein Field Embryo Anura Kiesecker & Blaustein Short Field Embryo Anura Anzalone et al Field Embryo Anura Blaustein et al Field Embryo Anura Blaustein et al Long Field Embryo Anura Hatch & Blaustein

103 Hyla versicolor Short Laboratory Embryo Anura Zaga et al Hyla versicolor Short Laboratory Larva Anura Zaga et al Litoria aurea 34 Short Field Embryo Anura van de Mortel & Buttemer 1996 Litoria aurea 34 Short Field Embryo Anura van de Mortel & Buttemer 1996 Litoria dentata 34 Long Field Embryo Anura van de Mortel & Buttemer 1996 Litoria dentata 34 Long Field Embryo Anura van de Mortel & Buttemer 1996 Litoria peronii 34 Long Field Embryo Anura van de Mortel & Buttemer 1996 Litoria peronii 34 Long Field Embryo Anura van de Mortel & Buttemer

104 Pseudacris crucifer 35 Short Laboratory Larva Anura Baud & Beck 2005 Pseudacris triseriata 35 Long Field Embryo Anura Starnes et al Rana arvalis 62 Long Field Embryo Anura Häkkinen et al Rana arvalis 62 Long Field Larva Anura Häkkinen et al Rana arvalis 63 Long 50 Laboratory Embryo Anura Pahkala et al Rana arvalis 63 Long 50 Field Embryo Anura Pahkala et al Rana aurora 44 Long Field Embryo Anura Blaustein et al Rana blairi 39 Long Field Embryo Anura Smith et al Rana blairi 35 Short Field Embryo Anura Bruner et al Rana blairi 35 Long Field Embryo Anura Bruner et al

105 Rana cascadae Field Embryo Anura Kiesecker & Blaustein 1995 Rana cascadae Field Embryo Anura Kiesecker & Blaustein 1995 Rana cascadae Field Embryo Anura Blaustein et al Rana cascadae Field Embryo Anura Blaustein et al Rana cascadae 44 Long Field Embryo Anura Belden et al Rana cascadae 44 Long Laboratory Larvae Anura Hatch & Blaustein 2002 Rana clamitans 45 Long Field Larvae Anura Tietge et al Rana clamitans 43 Long 14 Laboratory Metamorph Anura Grant & Licht 1995 Rana luteiventris 48 Short Field Embryo Anura Blaustein et al

106 Rana pipiens 49 Long Field Larva Anura Ankley et al Rana pipiens 49 Long Laboratory Larva Anura Ankley et al Rana pipiens 46 Long Laboratory Larva Anura Ankley et al Rana pipiens 45 Long Field Larva Anura Tietge et al Rana pipiens 37 Long Field Embryo Anura Long et al Rana pipiens 45 Long Laboratory Larva Anura Ankley et al Rana pretiosa 46 Long Field Embryo Anura Blaustein et al Rana pretiosa 47 Long Field Embryo Anura Blaustein et al Rana septentrionalis 45 Long Field Larva Anura Tietge et al Rana sylvatica 43 Short 2 Laboratory Embryo Anura Grant & Licht

107 Rana sylvatica 43 Long 66 Laboratory Larva Anura Grant & Licht 1995 Rana sylvatica 43 Long 14 Laboratory Larva Anura Grant & Licht 1995 Rana temporaria 62 Long Field Embryo Anura Häkkinen et al Rana temporaria 62 Long Field Larva Anura Häkkinen et al Rana temporaria 55 Long 22 Laboratory Embryo Anura Pahkala et al Rana temporaria 69 Long Field Embryo Anura Merilä et al Rana temporaria 59 Long Field Embryo Anura Pahkala et al Rana temporaria 66 Long Field Embryo Anura Pahkala et al Rana temporaria 63.5 Long Laboratory Embryo Anura Pahkala et al. 2002a Rana temporaria 55 Long Laboratory Embryo Anura Pahkala et al. 2002a 87

108 Rana temporaria 62.4 Long Laboratory Larva Anura Koponen & Kukkonen 2002 Rana temporaria 60 Long Field Embryo Anura Pahkala et al. 2002b Taricha torosa 34 Long Field Embryo Caudata Anzalone et al Triturus alpestris 47 Short Laboratory Larva Caudata Nagl & Hofer 1997 Xenopus laevis Short Laboratory Embryo Anura Zaga et al Xenopus laevis Short Laboratory Larva Anura Zaga et al

109 Table 3.2. Biological and methodological explanatory variables used in regression analysis of the effects of UVB on survival in amphibians. The asterisk denotes variables available for all studies. Variable Possible values Number of comparisons including the variable Latitude Continuous variable Duration of study Long (> 7 days) Short (< 7 days) 75 Days of UVB exposure Continuous variable 70 Elevation Continuous variable 79 Venue* Life history stage* Taxonomic order* Laboratory Field Embryo Larva Metamorphic Anura Caudata 89 (all) 89 (all) 89 (all)

110 90 Table 3.3. Additional stressors used in factorial meta-analysis Additional Life history Species Reference stressor stage Saprolegnia Embryo Bufo boreas Kiesecker & Blaustein (pathogenic Pseudacris regilla 1995 water mould) Rana cascadae ph Embryo R. arvalis Pahkala et al R. pipiens Long et al R. temporaria Pahkala et al. 2002b Larva R. cascadae Hatch & Blaustein 2002 Nitrate Embryo P. regilla Hatch & Blaustein 2003 Larva R. cascadae Hatch & Blaustein 2002 Landfill leachate Embryo R. blairi Bruner et al Copper Larva P. crucifer Baud & Beck 2005 Methoprene Larva R. pipiens Ankley et al Carbaryl Embryo Xenopus laevis Zaga et al Hyla versicolor Larva X. laevis H. versicolor Bisphenol A Larvae R. temporaria Koponen & Kukkonen 2002

111 91 Table 3.4. Final regression model after backwards selection. Variables were iteratively removed until only significant terms remained in the model. All explanatory variables were included (excluding studies that did not report all information in Table 3.1) or all studies were included (using a subset of variables indicated by an asterisk in Table 3.1). Factor df Type II F P All explanatory variables included Overall model (n = 58, residual df = 54, model r 2 = 0.414) Days of UVB Life history stage Taxonomic order All studies included Overall model (n = 89, residual df = 86, model r 2 = 0.182) Life history stage Taxonomic order SS

112 Table 3.5 Final regression models after excluding Blaustein laboratory studies. Variables were iteratively removed until only significant terms remained in the model. 92 Factor df Type II All explanatory variables included SS F P Overall model (n = 45, residual df = 41, model r 2 = ) Days of UVB Life history stage < Taxonomic order All studies included Overall model (n = 64, residual df = 61, model r 2 = ) Life history stage Taxonomic order

113 Other stressor treatment (ln) - 3 UVB treatments (ln) UV, +stressor -UV, +stressor 4 +UV, -stressor -UV, -stressor Calculations: Mean effects UVB exposure: (1 + 3) (2 + 4) Other stressor: (1 + 2) (3 + 4) Interaction: (1 3) (2 4) Individual effects UVB alone: (3-4) UVB in the presence of an additional stressor: (1-2) Stressor alone: (2-4) Stressor in the presence of UVB: (1-3) Figure 3.1. Conceptual framework for factorial meta-analysis. The value for the boxes (1-4) was the natural log of survival reported from each comparison. The mean effects were calculated as in Hawkes and Sullivan, 2001.

114 94 Figure 2. All studies, a a b ab b (89) (60) (25) (4) (78) (11) Life history stage Taxonomic order Figure 3.2. Mean effect sizes (log response ratio) for all studies included in the analysis of the effect of UVB on survival alone. All effect sizes are significantly different from zero except the metamorphic life history stage (M). Numbers in parentheses indicate sample size for each estimate. FM = full model, E = embryo, L = larva, M = metamorph, A = anura, C = caudata. Data are means and bias-corrected bootstrapped confidence intervals, except the confidence interval for M is a 95% parametric confidence interval. Means that share a common lowercase letter are not significantly different.

115 95 Mean effects Individual effects Figure 3.3. Mean effect size estimates for the factorial meta-analysis. Estimates are calculated as in Figure 3.1. Data are means and bias-corrected 95% confidence intervals. All effect estimates are significantly different from zero. U = UVB present, S = additional stressor present. Means that share a common lowercase letter are not significantly different.

116 96 Chapter 4 Larval amphibians seek warm temperatures and do not avoid harmful UVB radiation Betsy A. Bancroft, Nick J. Baker, Catherine L. Searle, Tiffany S. Garcia and Andrew R. Blaustein Submitted to Ecology

117 97 Abstract Habitat use by animals often reflects the balance between conflicting demands such as foraging and avoiding predation. Environmental stressors such as temperature can also affect habitat use in many organisms, particularly in ectothermic animals. Amphibian larvae in ephemeral ponds must grow and develop before their habitat dries or freezes. Warm, shallow thermal regimes in ponds can optimize growth and developmental rate but may also expose larvae to potentially harmful levels of ultraviolet-b (UVB) radiation. Exposure to UVB radiation may lead to mortality or result in sublethal effects, including physiological changes, anatomical deformities, and delayed growth. Thus, optimally, amphibians seeking sunlight for thermoregulation must balance this behavior with limiting their exposure to harmful UVB radiation. We conducted a series of laboratory and field experiments to test the hypothesis that larval amphibians avoid UVB by selecting microhabitats with lower exposure to UVB. We then quantified habitat use of the larvae of four amphibian species using field transects in three ponds with different UVB transmission. Larval Pseudacris regilla and Rana cascadae did not avoid UVB radiation in laboratory or field experiments and preferred warmer temperatures in laboratory thermal gradients, regardless of UVB exposure. In field surveys, the majority of anuran larvae were observed in water less than cm deep, whereas salamander larvae were most often observed in deeper, cooler water. The similarity in habitat use across different

118 98 sites and the lack of evidence of UVB avoidance in choice tests suggests that larval anuran amphibians may be exposed to high levels of UVB radiation due to habitat use. The observed habitat use may be the result of other pressures, such as thermoregulation, foraging, or predator avoidance. Introduction An individual s use of habitat can occur on a large scale (macrohabitat) or a small scale (microhabitat). Generally, the selection and use of habitat is assumed to have fitness consequences, as selecting and using advantageous or optimal habitats will increase the fitness of organisms within a population (Fretwell & Lucas, 1970; Jaenike & Holt, 1991). These habitat use decisions are most likely based on some set of environmental criteria. Both abiotic and biotic factors are responsible for the distribution of a population throughout a habitat and the resulting fitness outcomes to the organisms. Abiotic parameters such as light intensity, temperature, soil moisture and nutrient availability can influence habitat use and selection in many species (e.g., Huk & Kühne, 1999). Biotic factors such as food resources (prey), competition, facilitation, and predation can also influence the distribution of animals (Fretwell & Lucas, 1970; Kats et al., 1988; Rosenzweig, 1991). These abiotic and biotic factors can act as selection pressures, resulting in trade-offs between different habitat types or patches (Sih, 1980; Lima, 1998). Variation in both the abiotic and biotic environment can interact, resulting in trade-offs between physiological requirements (e.g., thermal

119 99 optima) and biological pressures (e.g., competition, predation) which often affect both micro- and macro-habitat use. Trade-offs between abiotic and biotic factors are especially important in ectothermic animals. Temperature greatly influences the behavior and ecology of ectotherms (Hutchison & Dupré, 1992), and many ectotherms can actively thermoregulate, using microhabitat variation to control body temperature (Huey & Slatkin, 1976, Hutchison & Dupré, 1992). However, thermoregulatory demands must be balanced with biological demands such as foraging, avoiding predation, and reproduction (Holomuzki, 1986; Downes & Shine, 1998; Martín, 2001; Martín & López, 2003). These trade-offs can result in complex spatial and temporal patterns of habitat use in ectothermic organisms. Trade-offs due to conflicting selection pressures may be particularly important in ectotherms with complex life cycles, such as amphibians. The early life history stages of many amphibian species develop in aquatic habitats of varying temporal stability (e.g., ephemeral ponds). Larvae in ephemeral ponds must develop quickly and undergo metamorphosis before the pond dries or freezes (Wilbur, 1980; Blaustein et al., 2001). Thermoregulation is particularly important for these amphibians, as growth rate is closely tied to temperature (Atlas, 1935; Lillywhite et al., 1973; Sype, 1974). Moreover, size at metamorphosis is related to adult size (Werner, 1986), and adult size is positively correlated with reproductive success (Semlitsch et al., 1988; Altwegg & Reyer, 2003). Shallow margins of ponds are generally warmer than deeper regions, but shallow regions frequently contain predators (Holomuzki, 1986). Larval

120 100 amphibians in ephemeral ponds are therefore exposed to a host of conflicting abiotic and biotic selection pressures. Amphibian populations are declining world-wide (Stuart et al., 2004), and a number of hypotheses have been advanced to explain the declines (Blaustein & Kiesecker, 2002; Collins & Storfer, 2003). One factor that may contribute to some of these population declines is increasing ultraviolet-b (UVB) radiation, which causes a variety of lethal and sublethal effects in many aquatic organisms (Bancroft et al., 2007), including amphibians (e.g., Blaustein et al., 1998; Pahkala et al., 2000; Häkkinen et al., 2001; Tietge et al., 2001; Blaustein et al., 2005). Sensitivity to UVB radiation may vary interspecifically, between life stages within a species, between populations and with environmental conditions (Blaustein & Belden, 2003). Exposure to ambient levels of UVB can increase mortality (Blaustein et al., 1998), reduce growth (Belden et al., 2000), alter behavior (Kats et al., 2000), and increase susceptibility to disease (Kiesecker & Blaustein, 1995; Kiesecker et al., 2001). Amphibians may avoid high UVB levels by seeking deeper waters (Belden et al., 2000; Blaustein & Belden, 2003; Licht, 2003). In choice experiments, larvae and adults of some species prefer areas with lower UV irradiance (Nagl & Hofer, 1997; van de Mortel & Buttemer, 1998; Belden et al., 2000; Garcia et al., 2004; Han et al., in press). However, avoiding UVB by seeking cooler, deeper water may reduce growth and create a conflict between thermal requirements and avoiding exposure to harmful levels of UVB.

121 101 Because of the accumulating evidence that UVB is harmful to amphibians, we conducted a series of experiments to test the hypothesis that larval amphibians avoid UVB radiation by seeking microhabitats with low UVB exposure. Thus, we 1) measured the preferred temperature of larvae in thermal gradients with a simultaneous UVB choice test, 2) conducted UVB choice tests in isothermic field enclosures to explore avoidance behaviors in the absence of a trade-off between UVB and temperature and 3) quantified habitat use of amphibian larvae by conducting transect surveys in different amphibian habitats. Materials and Methods Study sites: Our study sites were selected to include a range of pond-breeding amphibian habitats. We collected and sampled larvae in a large lake (Todd Lake, elevation 1875 m), a large ephemeral pond (Susan s Pond, elevation 1954 m) and an alpine meadow containing ~38 small ephemeral pools (Potholes, elevation 2300 m), all located in Deschutes County, Oregon (Appendix D). UVB avoidance in thermal gradients: These experiments tested if larval amphibians alter thermoregulatory behavior to avoid UVB radiation. Pseudacris regilla (Pacific treefrog) larvae were collected from the Potholes on 17 July 2005 and Bufo boreas (western toad) larvae were collected from Todd Lake on 22 July Larvae were returned to the laboratory and kept in 38 L glass aquaria filled with treated dechlorinated tap water. Animals

122 102 were fed a 3:1 mix of alfalfa pellets and fish flakes (TetraMin, Melle, Germany) ad libitum. To prevent acclimation to a constant temperature in the laboratory, all animals were kept at 15ºC from hr and warmed to 22ºC during daylight hours. All trials were completed within 10 days of field collection. Experimental lanes were created in 1.2 m aluminum rain gutters painted with Pratt & Lambert PalGard epoxy paint (Sherwin-Williams Company, Cleveland, Ohio, USA) and filled to a depth of 3.5 cm with dechlorinated tap water. Thermal gradients were created by filling a small metal pocket at one end of each lane with dry ice while the opposite end was placed on a hotplate. This method allowed the establishment of a ~10ºC difference between the two ends. Temperature was recorded throughout each trial with ibutton temperature loggers (Maxim Integrated Products, Sunnyvale, California, USA) placed every 27 cm along each experimental lane for a total of 6 ibuttons per lane. In addition, thermometers were placed at each end to directly monitor temperatures throughout the trials. Each lane was covered by two filters, one which blocks the passage of UVB radiation (Mylar) and one which permits the passage of UVB radiation (acetate). Each filter covered half the length of the lanes and two filters were placed such that the entire lane was covered by filters. During each trial, 4 lanes had the blocking filter placed over the warm end with the transmitting filter over the cool end, while the remaining 4 lanes had the transmitting filter over the warm end and the blocking filter over the cool end. Experimental lanes were placed under an array of UVB (Q-Panel UVB 313; Q-Panel, Cleveland, Ohio, USA) and full-spectrum bulbs (Vita-Light; Durotest Corporation, Fairfield, New

123 103 Jersey, USA), such that each lane was illuminated by both types of lights. UVB radiation was measured using a hand-held Solar Light meter with a UVB probe (meter model PMA2100, probe model 2102; Solar Light Company, Philadelphia, Pennsylvania, USA). In each experiment, UVB under acetate filters was approximately 10 μw/cm 2, while UVB under Mylar filters was virtually undetectable (< 0.05 μw/cm 2 ). These levels are within the range of ambient UVB levels in the Oregon Cascades (Belden et al., 2000; Kiesecker et al., 2001). Two sets of trials per day were conducted over three days for each species. For the first set of trials, observations began at approximately 1100 hr each day, while the second set of trials began at approximately 1400 hr each day. Our intent was to test UVB avoidance behavior of tadpoles during peak UVB hours ( hr). Once a stable gradient was established in each of the 8 lanes, tadpoles were randomly assigned to one of the two UVB treatments. Each unit contained one tadpole, for a total of 24 tadpoles in each treatment. Tadpoles were allowed to acclimate for 15 minutes prior to observation. After the acclimation period, the location of the tadpole in each gradient was recorded every 10 minutes for 100 minutes (10 observations per tadpole). The temperature closest to each tadpole for each observation was used to determine average temperature selected by each tadpole. If a tadpole was located equidistant between two probes, the average of the two temperatures was used as the preferred temperature. Field UVB choice experiments:

124 104 These experiments were designed to determine if amphibian larvae selectively avoided areas with higher UVB exposure in the absence of a thermal gradient. Trials were conducted between hr PDT on 22 July, 2005 (B. boreas, Todd Lake), 20 July, 2006 (R. cascadae, Susan s Pond), and 10 August 2006 (P. regilla, Potholes). Experimental units consisted of plastic boxes (34cm L X 21cm W X 11cm D) floated in the pond with Styrofoam floats on all four sides and mesh panels to allow for water circulation. One half of each container was randomly assigned a UVB blocking filter (Mylar). The other half of each container was covered with a UVB transmitting filter (acetate). Tadpoles (N=24) were placed singly in each box and allowed to acclimate for 15 minutes prior to observation. The location of each tadpole within the unit was recorded every 10 minutes for 120 minutes, for a total of 12 observations per tadpole. Halfway through the trials, each box was rotated 180º to avoid bias due to cardinal direction. After rotating the boxes, tadpoles were allowed 10 minutes to acclimate before resuming the trials. Temperatures on each side were recorded in a subsample of units at the end of each trial. UVB radiation was measured at the beginning of each trial under the filters using a hand-held Solar Light meter with UVB probe (see model information under Laboratory trials ). The Mylar filters transmitted approximately 10% of ambient UVB, and the acetate filters transmitted 75-80% of ambient UVB. UVB levels differed on each day, but were approximately 10-16μW/cm 2 under acetate filters. Field transects

125 105 We applied field transects at Susan s Pond, Todd Lake and the Potholes to quantify thermoregulatory behaviors and UVB exposure in natural systems. The distance between transects and length of each transect varied between locations based on water body size and shape (Appendix D). Transects were run perpendicular to shore and were divided into 1 m 2 sections. We attached vertical arrays of ibutton temperature loggers at 2m (Susan s Pond) or 3m (Todd Lake) intervals to record temperature at different depths. Each array consisted of ibuttons located at 20cm intervals from the surface to the bottom of the pond. At Susan s Pond and Todd Lake, an individual ibutton was placed in the shallowest portion of each transect next to shore. Individual ponds at the Potholes site were too small to contain more than one transect; thus, we selected four representative ponds and aligned each transect along the widest axis of each pond. The four ponds selected at the Potholes were physically different from each other and from both Susan s Pond and Todd Lake. With the exception of Pond C, the ponds were steep-sided without a shallow margin (Appendix E). Thus, ibuttons were placed approximately 1.5 and 3.5m from shore in two of the four ponds. All transects were in place by 1000 hr. We allowed sediment to settle for 30 minutes prior to walking transects. For each observation, two observers slowly walked either side of each transect, counting the number of larvae of any observed amphibian species within his or her half of the square meter. Care was taken not to disturb animals or sediment. We measured UVB radiation at the surface of the water and at every depth from 10 cm to 50 cm (or the bottom of the pond) before each

126 106 observation period using a hand-held light meter with a UVB probe (see model listed under UVB avoidance in thermal gradients ). Our goal was to capture larval movement due to diel thermal fluctuations. At Susan s Pond and Todd Lake, we walked each transect three times a day for two consecutive days. We walked the transects at Susan s Pond and Todd Lake in the morning ( hr), afternoon ( hr) and in the evening ( hr). Transects at the Potholes were in place for one day. We walked the transects at the Potholes twice, once in the afternoon (1200 hr) and once in the evening (1700 hr). The number of tadpoles observed in each transect varied widely across transects and at each sampling time. We standardized the data for each transect by using the total number of each species observed in each transect per sampling time to calculate the percent of each species observed in each square meter along each transect at each sampling time. We used Spearman s rank order correlation statistic to examine the relationship between the percentage of animals per square meter and the average depth of each square meter at each sampling time. Each day and time was analyzed separately. Results UVB avoidance in thermal gradients: Thermal gradients ranged from 20-31ºC in the Pacific treefrog trials and from 21-31ºC in the western toad trials. No difference in average temperature selected was detected between treatments in either P. regilla (p = 0.885) or B. boreas larvae (p =

127 ; Mann-Whitney U test; Table 4.1). Both species selected relatively warm temperatures (Table 4.1). Because we did not detect a difference in mean temperature selection, we used equivalency tests and calculated the least significant value using power analysis. The equivalency test and power analysis support our findings of no difference between mean temperatures selected in the two UVB treatments (Table 4.1). Field UVB choice experiments: We found no evidence for UVB avoidance in larvae of the three species in these experiments (binomial test; P. regilla, p = 0.16; B. boreas, p = 0.15; R. cascadae, p = 0.14). Three B. boreas larvae escaped during the trials and were excluded from the analysis. Several tadpoles from each species exhibited no choice between sides (i.e., equal number of observations on each side of the container; 5 B. boreas, 5 R. cascadae, 3 P. regilla) and were thus excluded from the analysis. No difference in temperature was observed between sides in any trial (Wilcoxon sign-rank test; B. boreas: p = 0.99, df = 5; R. cascadae: p = 0.13, df = 4; P. regilla: p = 0.99, df = 5). Field transects: Susan s Pond: Over the two days of observations in Susan s Pond, we counted 2151 A. macrodactylum, 984 R. cascadae larvae and 215 P. regilla larvae. These values do not necessarily reflect independent larvae; our methods did not include marking larvae and it is possible that the same larva could have been counted more than once.

128 108 However, within a transect and sampling time, we avoided counting individual larvae more than once. Water depth influenced both UVB penetration and temperature. The transects spanned a half-meter depth gradient (minimum of 3.1 cm to a maximum 51.3 cm). UVB penetration was relatively low in Susan s Pond: only 5.2% of surface UVB was detected at 10 cm (Figure 4.1). A negative correlation between depth and temperature was detected at all sampling times (Table 4.2). Greater thermal stratification was present during mid-day compared to morning or evening observations (Appendix F). Larvae were exposed to a maximum thermal gradient of 12ºC during mid-day (18º- 30ºC). Within the transects, water depth and temperature were correlated with species distribution. The relationship between water depth and larval distribution was different for each species (Table 4.3). The distribution of A. macrodactylum was positively correlated with depth at every sampling time except for the afternoon and evening of 19 July. At mid-day on both sampling days, the majority of A. macrodactylum larvae were observed in deeper water in temperatures at least 5ºC cooler than the maximum temperature (Figure 4.2a). In contrast, a negative correlation between the distribution of R. cascadae larvae and depth was detected during mid-day (1415 hr and 1430 hr). The relationship between R. cascadae distribution and depth was non-significant at all other sampling periods. Few P. regilla larvae were observed at 1015 hr on 20 July and 1800 hr on both days. Thus, we removed them from the analysis at these sampling times. A negative correlation between P. regilla

129 109 distribution and depth was detected at 1030 hr and 1415 hr. A marginally significant negative relationship was detected at 1430 hr. However, at 1430 hr, more than 80% of the larvae were found in water <5 cm deep, and the rest of the observed P. regilla larvae were nearly evenly dispersed across other depths. The majority of both R. cascadae and P. regilla larvae were observed in shallow water (<15 cm, R. cascadae; <10 cm, P. regilla) at temperatures within 2 ºC of the maximum temperature (Figure 4.2 b,c). Todd Lake: During the two-day survey at Todd Lake, we counted 3954 B. boreas, 185 R. cascadae larvae and only 3 P. regilla larvae. Because we observed very few P. regilla, we removed them from the analyses. At mid-day on both observation days, the majority of R. cascadae larvae (> 65%) were observed in water < 10 cm deep. R. cascadae were excluded from further analyses due to low sample sizes in each transect. As in Susan s Pond, water depth again influenced both UV penetration and temperature. Within the 7 m transect, water depth increased from a minimum of 3.3 cm to a maximum of cm. UVB penetration was much greater in Todd Lake compared to Susan s Pond (Figure 4.1). More than 60% of surface UVB was detected at 10cm. Strong diel fluctuations in thermal profiles were observed (Figure 4.3). In the early morning, a positive relationship between temperature and depth was detected, with deeper locations warmer than shallow regions ( hr, Spearman s rho = 0.47, p = ). Between hr, no relationship between

130 110 temperature and water depth was observed (Spearman s rho = 0.21, p = 0.15). After 1100 hr, shallow regions were warmer than deeper regions ( hr, Spearman s rho = -0.34, p = 0.002; , Spearman s rho = -0.46, p < ). This relationship persisted until 1800 hr, when no relationship between depth and temperature was detected ( hr, Spearman s rho = , p = 0.66). After 2000 hr, the shallow regions were again cooler than deeper regions ( hr, Spearman s rho = 0.47, p = 0.003). The maximum thermal gradient was observed at mid-day ( hr) with a difference of 8.5 ºC between the shallowest and deepest regions ( ºC). The distribution of B. boreas tadpoles within transects was correlated with temperature and depth. The direction of this relationship varied across sampling times and mirrored diel temperature fluctuations (Table 4.4). At the earliest observation (0945 hr, 10 August), more tadpoles were observed in deeper water (Figure 4.3a). At 1045 hr on 9 August, more animals were observed in shallow water (Table 4.4). More animals were also observed in shallow water during mid-day sampling on both days (Table 4.4, Figure 4.3). Indeed, at mid-day on the sunniest day (9 August), over 88% of animals observed were located in water less than 10 cm deep. More tadpoles were observed in deeper water during the evening on both sampling days (Table 4.4). On 10 August, we observed a large mass of tadpoles swimming towards the shore from hr. The tadpoles formed a thick band at the edge of the pond and appeared to forage in the shallows, occasionally stirring up sediment on the bottom of the lake

131 111 (Figure 4.4a). The tadpole aggregation was so dense that individual tadpoles had portions of their bodies out of the water (Figure 4.4b). Potholes: UVB transmission was greater in Ponds A and B than in Ponds C and D. In Ponds A and B, 30 % of surface UVB was detected at 10 cm, while in Ponds C and D, 8% of surface UVB was detected at 10 cm (Figure 4.1). The thermal environment in the Potholes was also different from the thermal regimes in Susan s Pond and Todd Lake. Larvae were exposed to a maximum thermal gradient of 5ºC (Pond B; Appendix G) or 3ºC (Pond C; Appendix G). In addition to structural differences between ponds, the ponds at Potholes varied biologically. Ponds A and D contained A. macrodactylum and P. regilla larvae. We observed A. macrodactylum, P. regilla, and R. cascadae larvae in Pond B, while Pond C contained larval P. regilla and R. cascadae. The physical differences between the ponds combined with the inability to run replicate transects in each pond made statistical comparisons difficult. The depth choices available to each larva were different between ponds, and the depth distribution of each species varied between ponds (Figure 4.5). To facilitate statistical comparisons, we ranked the depth classes (5 cm intervals) in each pond, such that the shallowest depth class was assigned a rank value of 1, the next depth class was assigned a rank of 2, and so forth and then used Spearman s rho to look for correlations at each observation time. Although this method obscured the details of depth within each transect, it allowed us to look for patterns in depth choice at the

132 112 scale of the whole site rather than individual pond scale. No correlations were detected between depth rank and percentage of animals observed for either A. macrodactylum or R. cascadae. However, a negative correlation between depth rank and percentage of P. regilla larvae was observed at noon (Spearman s ρ = -0.52; p = 0.016), suggesting that larval P. regilla were more likely to be found in shallower regions of each pond. Discussion: UVB radiation causes negative effects in all four species examined in this study. This includes increased mortality and a variety of sublethal effects including reduced growth (e.g., Blaustein et al., 1994; Blaustein et al., 1997; Belden et al., 2000; Blaustein et al., 2005). Therefore, the hypothesis that these species avoid UVB is reasonable. Accordingly, we used a combination of approaches to explore habitat use in amphibians, with specific reference to temperature and UVB radiation. In laboratory thermal gradients, larval P. regilla (Pacific treefrog) and B. boreas (western toad) did not avoid UVB by selecting cooler temperatures with lower UVB levels. Rather, larvae preferred relatively high temperatures, regardless of UVB exposure. In the field, larvae of these two species, in addition to R. cascadae larvae, showed no preference for low UVB areas in choice tests. Using similar methods, Belden et al. (2000) found that larval A. macrodactylum preferred shade to full sun but did not discriminate between differences in UVB levels manipulated by plastic filters. The results of the two experiments presented here, in conjunction with previous

133 113 research, suggest that the larvae of the four species in this study do not avoid UVB. These results were further supported by our field transects. If amphibian larvae avoid UVB in the field, we expect to see fewer larvae in shallow water at Todd Lake due to the relatively high UVB transmittance at this lake. Yet, we observed large numbers of tadpoles in shallow water (< 10 cm), despite high levels of UVB at these depths in Todd Lake. The distribution of larvae varied across sites, but anuran larvae (P. regilla and R. cascadae) were generally observed in shallow regions, especially during peak UV hours. In contrast, A. macrodactylum larvae were usually located in deeper regions, which is consistent with Belden et al. (2000). No pattern in depth choice was observed for either Cascades frog or long-toed salamander larvae at the Potholes. Pacific treefrog larvae were observed more often in shallow regions at the Potholes at noon. However, compared to the other two sites, the ponds at the Potholes were relatively isothermic at each sampling. Moreover, the distribution of larvae in the only Pothole with a shallow margin (Pond C) was similar to the distribution of larvae at Susan s Pond and Todd Lake: more than 70% of larvae were observed in less than 5 cm of water. Thus, larvae at these sites, particularly anurans, do not avoid sunlight and are exposed to relatively high levels of UVB radiation. The dose of UVB received by larvae is a function of many factors, including habitat use and the UVB transmittance of the habitat. The three sites used in this study vary widely in UVB transmittance. A recent study of 136 amphibian breeding habitats estimated a mean 40.3% percent transmittance of UVB at 10 cm for habitats

134 114 containing three of the four species observed in the current study (R. cascadae, B. boreas, A. macrodactylum; Palen et al., 2002). Both Susan s Pond (5.2%) and the four ponds at the Potholes site (29.6%, 26.0%, 7.9% and 3.1%) have much lower UVB transmittance at 10 cm than average. However, anuran larvae in our study were commonly observed in water < 10 cm deep, regardless of the UVB transmittance of each site. The sites in our study were not randomly selected but were selected to represent a range of amphibian habitats. Therefore, we cannot conclude that these sites are representative of all amphibian breeding habitats in the P acific Northwest. However, the similarity in habitat use among sites despite widely varying UVB environments suggests that, at the landscape scale, larval amphibians may be exposed to high levels of UVB radiation due to habitat use. Moreover, UVB did not influence habitat use in these sites. The observed habitat use may be due to other pressures such as thermal requirements, resource use, or predation risk. Behaviors such as thermoregulation, foraging, and predator avoidance commonly affect habitat use. Temperature is frequently the dominant physical factor affecting physiology and behavior of larval amphibians (Ultsch et al., 1999). The diel movements of B. boreas tadpoles at Todd Lake closely followed diel temperature fluctuations, suggesting a strong relationship between temperature and habitat use. Similar diel movements have been observed in the larvae of other species (Brattstrom, 1962; Beiswenger, 1977). Previous work on habitat use in B. boreas, R. cascadae and P. regilla suggests that temperature is the single most important habitat variable for predicting tadpole density (Brattstrom, 1962; O Hara, 1981). Wollmuth et al.

135 115 (1987) found the highest densities of Cascades frog tadpoles in the warmest (and shallowest) regions of a pond, and tadpole aggregations moved throughout the afternoon to track the warmest temperatures. Hokit and Blaustein (1997) observed aggregations of Cascades frog larvae in the warm shallow regions over potential food sources. Observations of larval Pacific treefrog aggregations suggest that they orient their bodies such that the maximal dorsal surface area is exposed to the sun, presumably for thermoregulation (Brattstrom & Warren, 1955). The importance of temperature as a cue guiding habitat use is not surprising given the influence of temperature on growth rate in amphibians (Atlas, 1935; Ryan, 1941; Álvarez & Nicieza, 2002). We did not quantify the distribution of potential predators or food resources in this study. Compared with the other species, habitat use in western toads is probably least likely to be affected by predators, as these toads have potent toxins in their skin (Arnold & Wassersug, 1978). Conversely, large salamander larvae and a variety of invertebrates eat Pacific treefrog and Cascades frog larvae (Peterson & Blaustein, 1992; Wildy et al., 1998) and these predators may have an impact on habitat use in these species. For example, the temperature where Pacific treefrog tadpoles were most dense at Susan s Pond was higher than the preferred temperature of this species in our laboratory experiments, suggesting that habitat use may be influenced by other factors such as the presence of predators as well as the thermal regime. The large long-toed salamander larvae were most common in the cooler deeper water in Susan s Pond; therefore, Pacific treefrog larvae could avoid predation by seeking warmer,

136 116 shallower water. In contrast, long-toed salamander larvae were not large enough to consume tadpoles in the Potholes, where little segregation between Pacific treefrogs and long-toed salamanders was observed. Habitat use in these larvae could be influenced by a number of factors not measured in the current study. However, UVB radiation does not appear to influence habitat use, despite the potential for significant damage caused by exposure to UVB. The effects of UVB radiation depend on the dose received by an individual organism, which is a direct consequence of habitat characteristics (i.e., UVB penetration, topography) and individual behavior. Some studies have suggested that habitat characteristics such as dissolved organic carbon may fully protect amphibians from damage caused by UVB radiation (Adams et al., 2001; Palen et al., 2002). However, these studies did not consider temporal and spatial habitat use within a pond and the potential for other selection pressures to affect exposure to UVB. Our results suggest that larval amphibians are exposed to potentially damaging doses of UVB as a consequence of thermoregulatory behaviors. Including the interplay between habitat characteristics and habitat use is vital to our understanding of how environmental stressors and conflicting selection pressures may affect organisms. Acknowledgements: M. Kavanaugh, L. E. Petes and R. H. Bancroft provided assistance in the field. Conversations with L. Crawshaw were essential to the design of the thermal gradients. B. A. Menge and E. Seabloom provided statistical advice. C. Baker provided housing

137 117 for BAB, NJB, and CLS during field observations. We thank B. A. Han, K. Kowalczyk, O. Fielding III and S. Poliakoff for assistance. Funding was provided by NSF grant (DBI ) to TSG, Oregon State University Department of Zoology Research Fund award to BAB.

138 118 Table 4.1. Results from Mann-Whitney U test, equivalency test and power analysis in thermal gradient laboratory trials. Treatment indicates the UVB exposure of the warm end of the gradient. Species Treatment Average temp selected df Z value P- value Equivalence p-value* Least significant value Bufo boreas UV shielded Pseudacris regilla UV exposed UV shielded UV exposed *Practical difference = 2ºC

139 Table 4.2. Depth and temperature correlations at Susan s Pond, Oregon on 19 July 2006 at each observation time. Time Spearman s rho p-value 10: < : < : <

140 Table 4.3. Correlation between water depth and mean percent animals observed for each species at Susan s Pond. Oregon. For P. regilla, fewer than 20 animals were observed at the observation times not shown. Ambystoma macrodactylum Rana cascadae Pseudacris regilla Time (hr) Date Spearman s rho p Spearman s rho p Spearman s rho p /20/ /19/ /20/ /19/ /20/ /19/

141 121 Table 4.4. Correlation between water depth and mean percent Bufo boreas observed at each sampling time in Todd Lake, Oregon. Date Time Spearman s rho p-value 10 August 9: August 10: August 14: < August 14: August 17: August 18:

142 122 Figure 4.1. UVB profiles from four Oregon study sites during mid-day. Todd Lake had the highest UVB transmittance. Only Ponds A and C are shown from the Potholes for clarity. The transmittance of Pond B was nearly identical to Pond A, while Pond D was similar to Pond C.

143 123 a b c Fig Distribution of the larvae of three species at mid-day on 20 July, 2006 in Susan s Pond, Oregon. Thick black line in the bottom panel represents the bottom of the pond and the numbers are temperature in degrees Celsius. Bars represent the mean percent larvae of each species observed at each depth + standard error (mean of three transects). The majority of anuran larvae (R. cascadae and P. regilla) were observed closer to shore in water < 15cm deep. Most salamander larvae (A. macrodactylum) were observed in water > 20cm deep.

144 124 Fig Distribution of Bufo boreas larvae in Todd Lake, Oregon, on 10 August The thick black line indicates the bottom of the lake. Contour lines represent thermal stratification in degrees Celsius. Bars indicate the mean percent (+ SE) B. boreas tadpoles observed in three transects in the morning (A), mid-day (B), and evening (C).

145 125 a b Figure 4.4. Toad tadpoles at Todd Lake, Oregon on 10 August 2006 (A). Panel B is a close-up view of the region indicated by the white rectangle in Panel A. White circles in panel B indicate tadpoles with at least a portion of their body out of water.